2014-06-07

Part A: Host Based Information Security and Security Operations in Military Digitized Environment

This is two fold paper on how to analyze defensive tactics in Global Digitized Environment. In this part A is introduced the military confrontation in digitized environment and aspects on improving host based ICT security.

Part B will include description of survivable management system structure and describe learning, training and exercising of Security Operation skills and understanding.

Introduction

Despite of high walls of defence (air gap, cross domain gateways, domain control, and physical access control) on the perimeters of digital structures, military is facing increasing challenges inside their ICT system of systems.  The opponent is increasingly breaching walls and effecting inside the fortifications via insiders (Manning), sub-providers (Snowden), air-gap-leaping viruses (Russia) and ‘waterholeing’, spearheading and phishing normal users. Militaries are defending their digital fortresses with more flexible defence, which requires continuous surveillance (monitoring), intelligent detection of incidents, eliminating unrelated events and rapid action in countering detected threats. Magnificent physical fortresses like Maginot-line or Atlantic Wall were breached or by-passed in history. Today military engineers of digital fortresses are building more flexible defence both within their digital walls. This is to gain more time to react and enable more variations of defence with modern defence-in-depth concept.

This paper is explaining current trends in implementation of host based security within digital military environment. The paper is also presenting means to improve understanding of interrelationships within ICT System of systems. The host based information security and flexible cyber defence is studied in environment of national territorial military defence. The strategic confrontational situation defined so that defender may achieve only partial dominance in any dimension of space of operations. The opponent is assumed to own abilities and to follow the art of operation typical to 2nd and 3rd generation of war  – industrial mass attraction or manoeuvrable forces enabled by network. Operations with other nature like peacekeeping, occupying or expeditionary and their possible 4th generation opponent  should have their own specific analysis.

This study is first defining confrontation in digitized environment to possible agendas of cyber offensive, targeting and vulnerabilities of defended System of systems. Secondly paper is describing implementation of a host based IT security.  Thirdly study is presenting a survivable structure for both physical and logical layers of security operations architecture. Fourthly there is definition of learning and training of human competence and skills to keep ahead of possible opponent.


1.Military Mission in Digital Environment


Military Confrontation

Military mission occurs most often when two or more parties of confrontation have drifted to conflict.  A simple model may be design to analyse basic interactions between parties and their environment as illustrated in figure 1.


Figure 1: A Simple model for analysing conflict between two military entities in digitized environment

Digitized environment is creating one additional dimension to normal military dimensions of land, air, sea and space. Digitized environment is creating complex interconnection between main opponents, their supporters and neutral parties. Digitized environment is making space of operation flat and global while reducing boundaries between crises and normal time and between fighting forces and other people. In digitized environment it is normal to use all possible means and avenues available to affect opponent. Friendly connected computers can be turned into attacking assets by turning them to remote controlled robots i.e. botnets. Opponent may utilize separate stake holders such as enthusiast, pro-movements or hired hackers to cover real sources of force. Russia was using their patriotic movement as a cover when attacking against Estonian governmental information assets.  Opponent may steal critical personnel and logistics information. They may jump over air gaps and block or change the situation awareness picture. Opponent may attack critical industries to block critical supplies. They may even do as simple as send emails to commanders to lessen trust to technology and reveal the weakness of military network.

Both parties are structuring their military force by utilizing System of systems enabled capabilities. These capabilities are using connections between different nodes to gain better situation awareness, quicker reaction, tailored effect and more sustained operation. System of systems at best is producing exponentially multiplied military effect  but at worst it may expose military to exponential or overconnectivity  vulnerability.
Military force is resembling the society that has provided it. Current 2nd and 3rd generation forces are connected to a large body of nonmilitary support. This logistic chain is essential to sustain force in operation. There are both information service providers, software developers and technical component providers for military information and communications technology systems. This important value chain might be utilized by the opponent to either gain access into military domain or to affect its role.
 

Issues of Offence

Potential opponent has wider variety of means in his reach in modern digitised environment than before, when he is facing a military force with system of systems structure. There are different targeting strategies shown in U.S. Gulf operation (Wardens  targeting strategy) and Russian operations in Estonia, Georgia and Ukraine. Here is an example from Russian strategy  of effect and operations with 2nd and 3rd generation force as depicted in figure 2.


Figure 2: An example of phasing of military operation evolved from Russian origins

Following Russian strategy there may be three overlapping phases of operation:
1. Projecting soft power and preparing for further measures,
Since using force does not mean waging war, there are many alternatives to project power and change both public, governmental and military behaviour before escalation to an armed conflict. Digitised environment provides number of means and avenues to collect information, to analyse structures of trust, to create targeting lists, to prepare space of operation and to effect on individual and collective opinion. This means destroying human trust in governance, public media and other people. In this phase more focused opponents have the luxury of persistence. They may operate with long goals, lessening risk of being detected and maximising effect when capturing valuable information and not bulk data. Recent sensation of NSA open-source intelligence gathering capabilities is just a proof that all nations are yielding digital sources.  Military Intelligence in Israel declared that they are under sustained cyber-attack. According to Maj.Gen Kochavi the cyber means are "the biggest revolution in warfare, more than gunpowder and the air power during the past century" . It took about one year from global society to discover STUXNET. One should wonder, how many malevolent software based advanced attacks or information gathering operations there are currently that has not been detected. Undetected attacks are the most dangerous in digitised military environment. There are no fingerprints, signature or behaviour that can be configured to intrusion detection systems.

2. Achieving strategic or operational surprise
Electronic shock means that opponent is using all avenues and means available to achieve strategic surprise by suppressing opponents System of systems function in early phase of conflict. Means to achieve this might include anything between high burst electromagnetic pulse to triggering planted worms to deny electrical utility services. The United States has been doing this but with kinetic means when they are operating with ‘Shock and Awe’ style to gain rapid dominance. The main aim of this style of operation is to deny all network enabled capabilities and force multipliers and destroy both cohesion of forces and their systems. The center of gravity of this offensive is the human trust to his systems and information. This happened when thousands of Iraqi commanders painfully learnt that their "closed-loop" private network was compromised. They received emails by the internal system from US Central Command demanding to leave tanks and armoured vehicles in formation and abandon them.

3. Destroying forces and their base.
Conventional tactics is using kinetic means to destroy both opponent’s force and their base for support to gain dominance for example at political level. This phase includes isolating physical area of operation from global network and focus effect on public opinions and behaviour. Kinetic effect means that missiles, special operation forces and land component are used to destroy key targets. Main aim is to suppress opponent’s control over his forces and destroy enough assets that force is shattered to dysfunctional components. Electronic warfare is used widely at tactical level during this phase.
The opponent is using means that are available in time and politics, penetrable through defence and effective in target. In digitised environment these means include electromagnetic, cyber, physical, insider, man-in-the-middle and psychological tools. Rational opponent is preparing his space of operation and trying to detect vulnerabilities from System of system of opposite side. These preparations will provide fast delivery of force and gain initiative by surprise, which is a main goal in any military operation.
Targeting

Rational opponent is analysing the System of systems of their possible adversary with operational and system analysis tools. Opponent seeks vulnerable relationships from interrelated structure such as trust between humans, information exchange, complex processing and physical buildings of key resources. This is illustrated in figure 3.


Figure 3: Interrelated System of systems as a target structure

Individual key people, human to human relationship and human trust to technology are the weakest links in information intensive System of systems structure. These links are opponent’s main targets.
1. Individual key people has following weaknesses that might be utilised by opponent:

  • Human being is egocentric in his thinking. People forget information or avoid to notice facts that do not support the adopted line of thinking or contradicts profound beliefs and values. Most famous example of this was the landing in Normandy, when German leadership was prone to expect landing in Calais, thus kept reserve forces at place too long to be effective.
  • People has also tendency to choose the path of least resistance. Information seeking individual uses the most convenient search method. The information seeking behaviour stops as soon as minimally acceptable results are found. People also have tendency to bypass any security control if that is the least effort way. 

2. Trust between people is also a fruitful target

  • It is hard to build trust between people, who do not know one another and are not working in closeness. Haphazardly organised task force or multinational, multiagency coalition has major challenges to align to one vision and cooperate with units they are not familiar with. This has been a major obstacle in NATO effort to reach Comprehensive Approach in ISAF operation.
  • Trust is lost easily in stress and it is impossible to renew in short time. Only shared awareness, persistence in practise and overcome difficulties in exercises may prepare people and their relations for these kind of challenges. Overemphasized unity of command will be one of the first targets for opponent to try to break up.

3. Information is being processed, stored and exchanged in digital format.

  • Information systems are so complex today, that it is difficult for individual or one control to define which action is normal (i.e. belongs to white list) and which is anomaly. Military information structure is volatile in time of crises. It needs special measures to be able to define between friendly and hostile.
  • Planted data gatherers may be passive and give only weak signs when they are sending captured information to their masters. Worms may change content in databases so incrementally that it does not show in daily queries or periodic reports. Only systematic comparison will reveal changes. The ability to run correlations through masses of information in one hand and high velocity of information in other hand are problem for military security operations.

4. Synchronised processing and presenting functions are special targets

  • Conventional multisensory tracking is suppressed by disabling synchronisation at datalink layer.
  • Manipulating of network timing will destroy the integrity of inputs and prevent monitoring based on time series analyses. 
  • Changing the thickness of processor circuit will degrade quality of pseudo-random generator thus weakening encryption algorithms. 
  • Injecting backdoors to whole fleet of network switches will enable to take down entire transport layer with one trigger signal that gets through perimeter defence. Only thorough testing of all installed devices and strategic distribution of sources will minimize these risks.

Most conventional way to disintegrate System of systems is to destroy them physically. This can be done by using both kinetic, explosive and non-kinetic warheads. A bullet through cable or antenna connection will cut transmission. One ordinary ballistic missile hit within 500 meters (CEP 500 m and warhead 800 kg) from unhardened communications facilities will destroy commercial of the shelf (COTS) devices. High powered electromagnetic pulse generated within couple of hundreds of meters from unshielded datacentre will destroy all integrated circuits. A special operation troop may breach and destroy a target per day and keep this pace up for two weeks behind enemy lines. Taking down general electricity or water distribution may disable whole digital infrastructure when normal backup services run out.

Topology of networked system is one of the main focus when analysing vulnerabilities. The aim is not to destroy all detected targets but those who give best disintegration as depicted in figure 4.


Figure 4: Topology as target

The rational opponent is not taking down targets as they appear in their sights but aim for more valuable attractions. For infantry these valued targets are leaders, antennas, engines, heavy weapon platforms, etc. Opponent is trying to disintegrate networked System of systems to pieces that are no more functional separately. One may think that 10 nodes of distribution and relatively good connectivity is enough to sustain services. If opponent understands the system structure, has means of effect available and can deliver them to targets, only four hits is needed to break the example network apart and most probably prevent service providing. Target list is even smaller if opponent can determine sensitive parts from ICT structure like synchronization hub, network management hub, network operation centres, access directories, home directories, database replication masters or master key managers. The center of gravity in network is not the number of nodes but centralised intelligence, logic and control of system.

A Sense of the Enemy

Empathy is intellectual identification i.e. ability to think like other human being. Sun Tzu said that if you know yourself but not the enemy, for every victory gained you will also suffer a defeat.   How does military gain knowledge of its potential opponents in digital environment?

The main task of Military Intelligence is to gather information of opponent’s resources, weapons, training, procedures, troops’ spirits and leaders’ behaviour. Intelligence is also gathering information of how opponent operates its forces in real situation. All this information is fused, analysed and used to build a model of opponent’s most probable behaviour in operation. This model is used as a basis for all enemy assessment, a skeleton for their force structure and a basis for anticipating their next movement. This legacy military intelligence process  is trying to define the most probable future ‘modi operandi’  of the opponent by collecting large quantity of information from past actions and then projecting it to situation in hand. In physical world major enablers or constraints are troop skills, level of gear, fighting spirit and culture of their leaders. Thus force cannot execute operation that is beyond its unified competence, strength of command and control and culture.

This is not necessary the case when opponent is operating in digitized environment. Troops are rather small teams that can learn quicker and change their method faster than massed forces. There is no direct fear of death involved so troop unity is not important. Even low level organization culture can produce major effects, if there are competent and motivated individuals within teams. Cyber attacking weapons are as good as their users are skilled to build them thus there is a race to harness most potential and motivated individuals to defend their national cyber structure. Only simple spamming, virus producing or Distributed Denial of Service (DDoS) attacks do not require long experience or advanced skills.

One way to understand and possibly anticipate opponent is to assume all the possible actions as he may be able to execute in area of operation. Knowing how to attack is a knowledge base for defensive operations. Thus two-sided exercises (BLUE – RED) in real or near-real structure are imperative. Rotation of personnel between RED and BLUE crews should also be systematic. The other way is to shape the area of operations and design it so that opponent has only limited courses of action to choose. One has to create digital pits, honeypots and obstacles to capture attacker’s behavioural pattern. After pattern recognition one has to seek pattern breaks that might tell about attacker’s weaknesses, intentions or base of force. Exploiting these will give defender an advantage or asymmetric position in confrontation.
A generic list of cyber-attack examples in military digitized environment is presented in Annex 1. 

2. TRUST BASED ON MONITORING AND REACTION CAPABILITIES FOR MILITARY ICT SECURITY


Concept of host based security

Military security architecture has been evolving from physical site based structure to domain based and from there to host based security.  Since number of interconnected domains became too big, trust relations too long and the size of singular domains reached tens of thousands users, it became evident that perimeter based IT security fortifications were not sufficient. There was something to be done behind high walls of domain perimeters to control security and react when malevolent incidents were detected. This remains of basic military tactics to survey larger space of operation and react with tactically positioned rapid reaction forces to counter any detected insurgency.

Host Based Security System , HBSS was created in US DISA originally with McAfee and BAE Systems, who were given the contract to implement first monitor agents to all hosts in network. Hosts that either process data or store it are provided with firewall and application level blocking. After creating these mini fortifications from each host, DISA went to install set of agents or equal function that are installed to hosts and communication nodes to monitor events and produce required surveillance data to be forwarded to Security Operation Centre (SOC). SOC has control over IT Operations change management and configuration management and can quickly change system configuration, when malicious behaviour is detected from information environment as illustrated in figure 5.


Figure 5 an example of Host Based Security concept

As in physical space of operation this creates a continuous competition of measure – countermeasure – counter-countermeasure. This symmetric confrontation is not end state for military operational art but asymmetry is required.

Detection of anomalies

The first basic principle was to define ‘red list’ of known malevolent events that were configured into detectors to give alert when detected. Later the number of possible anomalies become too big to follow and ‘whitelist’ was created to baseline the authorized events in hosts and networks. Everything deviating from ‘whitelist’ was categorized as security incident. Then opponent become more professional and had time to be more persistent – Advanced Persistent Threats, APT . Detecting anomalies was done by comparing existing events either to baseline of authorized behaviour or to vectors of know malevolent attacks. This required capturing more events (Full Packet Capture) and storing them for longer periods (Big Data). Then complex task of correlating different events with time series, graphical presentation or statistical analysis methods was executed. Actual sense making requires highly skilled, motivated and experience individuals, which are expensive to train (min. 5 years) and hard to keep (current hype in cyber defence markets) within military forces. Collecting data for longer periods created a big data problem, which is being solved by virtualized information management and utilizing for example HADOOP  data-set processing structure. Some examples of setting threshold are illustrated in following figure 6.


Figure 6: Examples of setting threshold to detect different attack vectors

One may define normal baseline, set detection of anything outside of this content, protocol, size, time or user behaviour. One may also assess vulnerabilities after patching and set threshold to detect anything that is trying to exploit the remaining vulnerabilities. One may define normal behaviour between clients, network, application server and database server. Threshold is set to detect any anomaly outside this behaviour. A continuous tuning cycle was created when new threat was introduced, secure design was found by testing, configuration changes of ICT structure was implemented and detection thresholds were reset. This struggle is coming less and less effective to defender since within first quarter 2014 only there was record high 15 million new samples of malware created.  More than 160 000 new malicious specimens per day like trojans, worms, viruses and adware are overwhelming any security organization if detection is based on analysed fingerprints of each version.

To meet this velocity and versatility of threat in digital environment, defenders have initiated following measures:

  • Ecosystems has been created around detecting malevolent software threats, there are both private and public laboratories that are sharing information of both fingerprints of malware and signatures of attackers.
  • National CERT or Cyber Defence Centres have created networks to share information on vulnerabilities, attacks and malevolent tools. European Network and Information Security Agency, ENISA has collected European Union national CERTs to share technical information. 
  • There are some taxonomy work ongoing between governments and cyber defence industry to help structuring information and sharing it. U.S. Department of Homeland Security sponsored Structured Threat Information eXpression (STIX)  and its exchange format TAXII are widely developed and used. ITU-T is promoting their CYBEX standardization work.  IETF has two standards called IODEF and RID to support the exchange of information.
  • With collected big data one may improve analyses and gain ability to detect more advanced and persistent attack vectors, behavioural anomalies of possible insiders, etc.
  • With strategic system management military organization may achieve multilayer approach to malware detection. Desktop based detection is done with one product based for example fingerprints. Server based detection is done with second product also based on fingerprints but independently created from first vendor. Gateway level protection is done with third product, which might be more heuristic based detection. Heuristic may be based for example on run time sandboxing, file and function analyses or virus DNA detection. 

 

Reaction measures

When anomaly is detected, alert is forwarded to SOC, which creates security incident and starts executing countermeasures. Countermeasures include several configuration changes in IT Operations thus the process integration between security and Information Technology Service Management, ITSM is imperative for speed of reaction. Within military Network Operation Centre it is an advantage if security, communications and IT-services operation processes are as integrated as possible to ensure better awareness, quick reaction and integrity of information over all administrators. Thus it is normal that there is only one incident management process for all incidents to get widest awareness and quickest reaction for each incident. Change management is also integrated between ITSM and security operations since any change in ICT structure might also require change in thresholds and white lists. Service support (HelpDesk) is used to deliver information on effects of incident to users to further the recovery with backup services. Figure 7 is depicting this process integration. Reactive approach is further integrated for example by Booz and Co who call their integration as Dynamic Defence Approach to Cybersecurity.


Figure 7: Example of Security and ITSM processes integration in military NOC

Proactive measures

Since reactive defence is always weaker than attack, a proactive arm was created. Security engineers are using Sandbox environment (referred as REFNET in figure 7) to maintain small copy of operational ICT system. Within this reference environment engineers are testing vulnerabilities of each ICT component, subsystem and System of systems. Testing includes scanning known vulnerabilities, penetration testing with variety of tools, fuzz testing  to find unknown zero-day  vulnerabilities and red team attacking against whole System of systems with variety of attack vectors.

ICT components are patched and hardened before they are delivered to operations environment (MILNET) by IT Operations change, configuration and release management processes (see figure 7). This ongoing process keeps Operational ICT up-to-date and as sustainable as possible. In ICT systems, there are always known vulnerabilities, which cannot be remedied. To compensate these vulnerabilities, special thresholds could be installed around most vulnerable components to detect any attempts of exploitation as explained in figure 6.

Normally this proactive arm is integrated with risk management or presently in concept of Governance, Risk management and Compliance (GRC) processes. In military environment these processes are amalgamated to operation security processes of 3 branch, military intelligence processes of 2 branch and C5I processes of 6 branch for example in US DoD structure. These processes may present orientation to tactical SOC and NOC processes like:

  • Value of assets to current mission by OPSEC
  • Information requirements by MI
  • Service protection requirements by C4ISR and
  • Information Operation tasks by Cyber.

 

Intelligence in Cyber space

To avoid the syndrome of island defence and to extend information gathering to further area of interest, there is a need for even an isolated intranet defence to be connected to global world of Internet. Internet can host “honey pots”, which lure possible adversaries reconnoitre actions and exploitation attempts. Honey pots are used to capture first-hand attack vectors and malevolent behavioural features as depicted in figure 8.


Figure 8: An example of Security Operations Centre information management within Host Based Security

There is also a need to be connected with larger society of cyber experts that continuously monitor malevolent behaviour in networks, collaborate with them and make better sense out of incidents in networks. This is basic intelligence gathering from surrounding environment. There are number of societies that provide both vulnerability information (like CERT or NSA organizations), event data (like McAfee recent threats), attack profiles, vectors and signatures (like CyberISE) and information on attackers themselves (like SLIC threat feed, Arbor Networks, VeriSign iDefense). There is also major laboratories and SOC service companies that provide restricted services (like LocheedMartin, BAE Systems, Thales, Northrop Grumman, IBM, HP, RSA, Nokia, Ericsson). This information is analysed, modelled and transferred to scripts for anomaly detection agents in Operational Environment. Security Information and Event Management (SIEM)  is gathering all alerts and events from detection agents and logs, correlating them in near-real-time and presenting picture of possible malevolent events within IT Operations environment. This is typical multisensory information fusion based military surveillance function. Typical challenges are: what is normal and abnormal, are analysts able to detect all malevolent incidents from mass of anomaly events, what if opponent can model the fusion method and can bypass it, what if opponent attacks against monitoring system itself and how to normalize data from different sensors to be able to detect same target?

There are several ways to improve detection probability and minimize effects of overload. So called Business Intelligence and Big Data methods  are enabling correlation of information received from different sources.

  • One can combine both structured and unstructured data extracted from machine level transactions (like IP packets), actual content (like email writing), context information (like time, role, place), session information (relations between instances), process information (relations between functions). After combining and normalizing one should correlate data to first understand what is normal and then be able to detect abnormality. It requires complex data management to be able to correlate data from relational databases, nonSQL data stores, distributed HADOOP structures and HTML pages even near-real time.
  • After data has been retrieved, there comes the most vulnerable phase of data reduction, normalizing and classifying. This is creating the basement for analyses and also the very information opponent might be using against defence if allowed.
  • Big Data can be analysed with simple graphical tools at presentation layer that shows distribution of events from several dimensions, amount of events relative to other categories, amount of hits per time, tangents, integrals, curve fits and interpolations.
  • Statistical analysis can be done for example over two sets of samples: behavioural data from past and present. While comparing these datasets there is a possibility to detect change and that change can be defined either as anomaly or evolutionary change of normal behaviour.
  • Modelling is using some logical language to describe the behaviour of a defended system or offensive system. Modelling enables to play or simulate with different scenarios while changing variables. If model differs from real world, then it becomes master instead of slave like it has happened with financial models . There are several different algorithms and methods that can be used to model a phenomena. 
  • Predictive analysis is based on assumption that one observed behaviour is similar to another. There are algorithms like recommenders, classifiers and clusters that try to predict future based on probability. 
  • Clustering like k-means algorithm  takes large set of data sets, arranges them with different clustering techniques and shows visual similarities. More advance way is to use several dimensions in cluster model like Bearing Points HyperCube.
  • Classifier like k-nearest neighbour algorithm is assigning class to object whose class is unknown.  It is imperative to choose k tailored to each environment of data retrieved.
  • Recommender algorithm is measuring distances between two data items and recommending based on vicinity. Data objects that are closest to one another tend to have same preferences. 

As attacks are coming more advanced and persistent these analysis methods can be utilized to capture unknown attack profiles or detect silent signs of possible scenarios.

While number of hosts increases, their interrelated behaviour evolves faster, transmitted information increases or attacker’s skills improve, host based surveillance will produce more and more data. It will become more challenging to define malevolent events from this big data. This requires very skilled analysts that are committed to improve their skills indefinitely. There is a risk that attacker remains stronger in this confrontation, because he has time and tools in his side. Protected system and surveillance should provide more time to detect and react. To counter this basic military concept of defence in depth has been applied also in cyber space.

Defence in Depth

There is the basic strategy of defence in depth  that has been used in military operations and now it is being utilised also in ICT security and cyber defence . One can build depth in multitude ways:

  • By modelling opponent’s behaviour in particular space of operations, one can detect and prioritize countermeasures according to opponent’s lethality and aims. LocheedMartin defined their Kill Chain  concept and published it 2011. Since then and parallel many de facto practises have included similar profiles for advanced threats.
  • Information and Communications Technology provide multiple layers that can be utilized wither to tailor detection per layer or create defensive structure on one particular layer. In very basic model one can define four layers of endpoint, access, application and storage. ISO OSI layer structure provide model for both defence and attack  in cyber space. There are specific strategies like Moving Target Defense (MTD) by Coronado Group that is using Self-Cleaning Intrusion Tolerant (SCIT) technology. 
  • Protected system itself provides depth similar to the terrain for Land Component. Skilled defence modifies the structure of their protected system to better enable monitoring and executing of countermeasures. These domains may include surrounding domain that enables to monitor opponent in action. Access domain where all users (both friendly, neutral and unfriendly) are forced to security scrutiny. Presentation domain where all sessions are checked and identities are connected to roles. Process domain defence is tailored to manage the behaviour of transaction and service query. Storage domain is the final line of defence against losing integrity and availability of data. There data may be encrypted both while resting and being processed.

Following picture 9 is illustrating how combining these three depths of attack behaviour, technology and domain structure, defence can achieve a remarkable advantage over any attacker from both inside and outside.


Figure 9: An example of Defence in Depth concept in cyber space

Military cyber defence may have strong perimeter based defence (like Nato mission networks in Kosovo) or face world that interconnected systems are not any more separable (like Future Mission Networks). It is possible to build defensive perimeters inside each other by using both technical layers and domain layers of ICT structure that can be either self-cleaning or agile when implemented with Software Defined Infrastructure or Software Defined Security as defined by VMware salesmen . When this is combined with active intelligence, surveillance and active probing of surrounding domains, one can build agile defence with time to act and provide several means to counter attacks.

Air gap has been one of the most powerful domain perimeters. It is still preventing basic attacks but not advanced attacks that can hop over air gaps using for example traditional data media transfer, injected features or alternative communication means . Air gap is also creating a lot of human based data transfer which might be even more insecure that using well build and controlled cross domain gateway. Although military is using a lot of air gap isolation their security arm should extend their intelligence and surveillance far over these air gap cyber walls. Even Genghis khan’s nomadic force was able to take over well advanced Chinese fortification of Zhongdu (modern Beijing) in 1215.


1 Annex: Threat scenarios


Estimated threat scenarios

With security operation functions implemented protected intrasystem should be able to stand following threat scenarios without availability loss more than 99%, maintain integrity of data with 95% of data asset and prevent higher classified information breaches totally.

Normal malevolent intents and user mistakes

(1) End user brings malevolent software from other sources via memory or some other device with installed programs and information. Either end user connects memory to terminal PC via USB or Bluetooth or end user connects device directly to network. This may cause cases like antivirus failed to clean, excessive scan timeouts from antivirus, scanning or probing during unauthorized time window, anomaly in suspicious activity baselines, anomaly in network baselines, anomaly in application baselines, etc.
(2) End user inputs incorrect data or executes some operation that violates information integrity. This may cause cases like anomaly in network baselines, anomaly in application baselines, anomaly in database query baselines, multiple logins from one location, device out of compliance, etc.
(3) End user downloads information to attached medium/device and takes it away (classical Manning case). This may cause cases like anomaly in suspicious activity baselines, unauthorized user access to confidential data, unauthorized device on the network, etc. Manning case is classical example of this.
(4) Admin installs malevolent program or makes mistake in installation. This may cause cases like excessive traffic on network, process failed to work, service account denied, anomaly in suspicious activity baselines, anomaly in application baselines, logging source stopped logging, etc.
(5) Admin gives important information to outsiders or insiders to utilize in security breach. This may cause cases like logs deleted from source, excessive exploit traffic, excessive traffic inbound (if gateway), anomaly in network traffic baselines, anomaly in application baselines, etc. A Classical social engineering by Mitnick is an example.
(6) Outsider gains access through cross domain gateway and injects programs that either collects data or violates its integrity. This may be seen as SMTP traffic from unauthorized host, excessive http traffic outbound, excessive exploit from single source, etc. A basic advanced persistence threat (APT) is typical example.


Generic opponent in preconflict situation

Before actual conflict opponent is utilizing all time to collect useful information either to capture better understanding of military decision making, force and planned operations or to prepare attack against information assets and services. This information collection is done in utmost secrecy because becoming revealed ignites countermeasures at other lines of operation.
Opponent has long time to prepare and resources in use but secrecy requires utmost clandestine approach. Opponent is also preparing computer network attack against system by installing countermeasures either against data integrity or service availability.
(7) Most cost-effective way would be providing devices with already installed information collection services. These malevolent features are often installed in operation system level, storage-programs or printer/scanner programs because it is within their normal function. Connection outside may be arranged by wireless or utilizing electricity system. Best location for data collection would be in data center, where no one person often recognises all devices or connections. Best way to manage installation would be either via vendor support or paid administrator. This may be seen as anomaly in application baselines, anomaly in SQL baselines, anomaly in network traffic baselines, anomaly in maintenance procedures, caps in logging, underperformance in devices or it will appear in vulnerability or fuzz testing. Classical example is US claims that there is unwanted features in Huawei switches.
(8) Special data collectors may also be installed in system where for example relational database is easily reachable via unmonitored SQL-connection. Data retrieval may be arranged via regular maintenance visit to premises. This may be seen as anomaly in network baselines, anomaly in SQL baselines, changes in database content, anomaly in backup procedures, etc. Data collectors are like network tap  or packet analysers .
(9) If information systems have uncontrolled connections outside, a simple data collection software may be utilized. Flame was one of this kind intelligence gathering virus. This may be seen as excessive traffic outbound, service account access to outside, anomaly in network baselines, etc.
(10) Disposed storage mediums are classical data collection targets. It is almost impossible to wipe data out from electromagnetic disk or memory for ever. Collecting used disks, circuits and tapes is very cost-effective. Hard disks of printers tend to be most neglected items when disposed. This may be seen in audits.
(11) Attack preparations are done most effectively by leaving backdoors or installing triggered scripts to devices when they are being delivered to target system. Using these backdoors is another issue. This may be seen as abnormal performance difference, in vulnerability scanning and in reference testing with overload etc.
(12) Network worms and other programmable agents that migrate in networks and operation systems is another cost-effective way to prepare attack. They are easy to install in closed networks by paid end user or visiting maintenance. Triggering is another issue. This may be seen as anomaly in configuration baselines, anomaly in process baselines, anomaly in network baselines, problems in installations, etc. FLAME software is classical example of this.
(13) Preparing and manipulating key administrator of protected system is classical measure. Money and all vices are used to prepare person so that in the brink of attack he/she executes triggering or disables services within admin rights. This may be seen in security surveillance of personnel background, anomaly in social network, anomaly in email traffic, anomaly in physical access, etc. Classical infiltrations like Kim Philby, Ray Mawby or Vilho Pentikainen are examples for this.
(14) Normal user session downloads a malevolent application from Trojan web site. Malevolent application hides itself into MS server structure and starts capturing specific data. After collection is connecting to FTP-server or SMTP-server and sending out coded content in small portions. This may be seen as excessive traffic outbound, service account access to outside, anomaly in network baselines, etc. Spearfishing and waterhole attacks are classical examples.


Generic opponent in the brink of conflict

Just before attack or conflict is the best time to disable information systems or change data content massively. This creates a shock effect through all network users and people trust to information systems or their content is lost for long time. Digitized force may be paralyzed totally since their system of systems is not working any more. This happened example early in both Iraq operations, when centralized air defence system was disconnected as separate sensors and weapons unable to function together. Cyber-attack is also very possible in situation when disabling digitized force opens new possibilities for pressure at political level. Targets are usually amongst most important databases or registries, hubs of trust structures or management system.
(15) A script in ERM-system is triggered by paid admin and gradually changes all HR and Material information. This is seen as excessive DB activity, excessive application load, excessive network traffic, anomaly in storage load, anomaly in back up service, etc.
(16) Information in user registry is changed from root level downwards. This is especially effective in centralized trust structures of Single-Sign-On service. This may be seen as anomaly in any key registry, anomaly in replication procedure, anomaly in PKI service, anomaly in HR registry, etc.
(17) Erasing private keys from PKI root disables all related identification and encryption services. This is seen as denial of PKI service, anomalies in normal encryption services, widely noticed denial of access, etc,
(18) Uploading malevolent code within normal update of programs or even within fingerprint file for virus detection as is told to happen in first Gulf war to US troops. This may be seen as excessive scan timeouts from antivirus, tests in reference environment, etc.
(19) Management system is quite often least protected system since IT professionals do not want to restrict themselves. Disabling monitor system may be done either by emptying configuration items MIB’s or by destroying asset management information in both active and passive databases. Disabling often centralized control servers dismantles network operation center’s ability to do changes in coordination. This may be seen as loss of management connections, anomalies in management system performance, anomalies in management network traffic, excessive management traffic in-band, etc.
(20) As in any sensor system, opponent may also overload security monitors with number of cut’s in IT-systems and utilize this red screen time to do something more lethal in systems. This is seen as multiple alerts coming to both NOC and SOC monitors, electricity cuts, excessive booting activity, etc. A Classical Distributed Denial of Service (DDoS) creates overflow of security alerts. This is also happening in event of major power cut and major node destruction.


2014-04-21

A Short Story of Military IT-security evolution and Map of Roads it has left behind

WHAT IS MILITARY IT-SECURITY?


Information Technology Security (IT Security) is information security applied to computers and computer networks. Information security is the practice of defending information from unauthorized access, use, disclosure, disruption, modification, perusal, inspection, recording or destruction. Information in this case includes all its forms from data to knowledge. All this can be defined as it is done in ISO 27000:2013: “Preservation of confidentiality, integrity and availability of information whilst taking care of authenticity, accountability, non-repudiation and reliability”. 

Information has to have these dimensions of security quality as illustrated in fig 1:

Confidentiality, which is preventing the disclosure of information to unauthorized entities.
  • Data confidentiality is preserving authorized restrictions on data access and disclosure.
  • Privacy assures that individual control of influence what information related to them may be collected and stored and by whom and to whom that information may be disposed
  • Authenticity is the confidence in the validity of a transmission, a message or a message originator

Integrity, which is the accuracy and consistency of information thorough its entire life-cycle.

  • Data integrity assures that information and programs are changed only in a specific and authorised manner.
  • System integrity assures that system performs its intended functions in an unimpaired manner.
  • Accountability assures that actions of an entity can be traced uniquely to that entity. 

Availability, which means that information and its processing services are ready for and in use of authorized users when they so require.


Figure 1: Classical structure for IT security

IT security is built with layered structures of technical (hardware, software and communications) security, physical security, personal security and organizational security measures.

This paper does not refer to embedded computing but to military utilization of commercial computing devices and their security measures. Thus technology roadmap begins not from Polish Bombe machine that they used to decrypt German Enigma coding but with mainframe computing (1964 IBM System/360 was market leader ) utilized for military resource management, statistical calculations and human resource management . From this on military IT security has tried to sustain mainly confidentiality and keep up with exponentially increasing technological complexity of computing and datacommunications.

WHAT IS IT SECURITY BASED ON?

Information management has been passed on printed mediums since Gutenberg revolutionized the management of explicit knowledge about 1455 by printing first Bibles. Before that information was mainly tacit. Cambridge University Library had 1424 only 122 books . Gutenberg changed totally information management procedures and at height of this era when 2009 the number of books in English in Nielsen BookData catalogue was over 8 million .
Books has been stored in library and their reading has been controlled by librarians since famous libraries of Nineveh and Alexandria. This has created the very basis for information management and thus information security. Books, their publication and copyright system has taken care of information integrity. Books have been available in libraries and librarians have controlled who were reading them. 

Figure 2: Gutenberg, library and human tendencies from information security point of view

Human beings are naturally curious. Through exploration and investigation they learn from their discoveries. They also have tendency to collect items like books sometimes over exceeding their actual need to read them. To counter this behaviour confidentiality has based on restricting people to get their hands on books that contain sensitive or unapproved information. Library has been physical site where both confidentiality and availability has been maintained. Books are available but their reading is under control.

People has also tendency to choose the path of least resistance. This is specially studied in information science. Information seeking client will tend to use the most convenient search method. Information seeking behaviour stops as soon as minimally acceptable results are found. Thus people have tendency of bypassing any security control if that demands fewer effort than following security procedures. 

The history of information management is based on physical structures of libraries that constraint users access to information managed in books, shelves, catalogues and bibliography. Librarians have been in key roles to project security policy by creating bibliographies of books and their authors, storing them in shelves, rows and rooms with classification system and further logging books and their loaners.  This was the legacy structure that has defined the early years of information technology security measures.

TRUST ON PHYSICAL STRUCTURES

Very first Mainframe computers were installed in closed rooms sheltered preferably underground to create multi-layered physical security specially to control access of human beings into computer services and digital information. Human beings and printed information were considered main avenues for information breach thus computerizing was not considered a specific risk as illustrated in fig 3.


Figure 3: An example of classical site structure based security trust 

Data communications was not considered since information feed to mainframe and outcome was mainly via punched cards. Later when terminals were introduced it was estimated that risk of someone tapping special wiring with right code breaker was non-existent.

Printed documentation was categorised, labelled and kept within buildings and inside vaults. Problem occurred when documents were transferred between sites. This risk was countered with online data communications encryption devices which were evolved from history of online communications security starting from Green Hornet during WWII. Printed documents from trash bins, mail and couriers were main points of intrusion for outsider to get documented information. Most information was captured by insiders  or with evolving telecommunication technology through social engineering  and telephones. Human Intelligence was adversaries’ main method.

During 1985 studies relieved that computers were transferring unintended electromagnetic waves which could have been received outside of trusted physical site. To control emission the TEMPEST standards (a minimum of 100 dB insertion loss from 1 KHz to 10 GHz) were invented and computers were installed inside Faraday shield. All electromagnetic emission is cancelled in conductive wall material of this structure (usually metal) and no emission is leaked through cage. 

There are many different categories for physical security but one generic is as depicted in fig 4:
  1. Level perimeter: required to process top secret information
  2. Level perimeter: required to process secret information
  3. Level perimeter: required to process confidential information
  4. Level perimeter: required to process restricted information


Figure 4: Generic classification of physical security in military compounds

These perimeters are normally built within each other so outer shield protects inner. In typical military compound outer fence with access control by gates and either security personnel or electronic systems based surveillance provides 4th level perimeter. 3rd level security perimeter is achieved when inside 4th perimeter there is building or shelter that has separate access control and structure that prevents other avenues of access. Consequentially 2nd and 1st perimeters are inside 3rd with separate access management and additional protective structures. During peace time these structures may be design to prevent physical breach until reactionary forces arrives to capture intruders. When war time structures are designed, survivability under attack is the greater goal.

Still one finds that computers in within 1st perimeter are often offline and accessible only to personnel who have physical access to top secret facilities. These computers are unfortunately weak when it comes to resist cyber threats over air cap as it has been demonstrated in cases of Stuxnet and Agent.ptz (latety version SNAKE). 

TRUST BASED ON CONTROLLABLE DOMAIN PERIMETER

As information management includes dissemination of information at least need-to-know bases, there was the need to connect secure physical military sites to other similar. This was established by having improved communication security on the lines that were used between computes in physical shelters as defined in fig 5.


Figure 5: An example of IT security that is based on trusted domains

Encryption started at very low level of datalink layer 2 or similar within line modem devices. With Ethernet protocol at Local Area Network and several Wide Area Network link structures encryption struggled to provide both performance and cryptography required. Digital COMSEC at layer 2 has evolved from preISDN 64 kpbs online devices and 2 Mbps PDH trunk encryption devices to current GigaEthernet encryption solutions that are either embedded into switches and routers or build in separate devices. Since physical security was structured as onion, also COMSEC was defined to follow the familiar one inside the other -structure. If layer 2 encryption did equal perimeter 4, then additional layer 3 encryption was good for perimeter 3 and additional layer 4 encryption was equalling perimeter 2 domain. 

So more and more data transfer capability was utilized to encrypt IP-packets at different layers. There was not much sense making utilized on how much risks really was mitigated with this layered encryption but it was utilized since it was available and familiar way to solve security problems. One example of this design is illustrated in fig 6.


Figure 6: An example of COMSEC architecture in domain based IT security

Here basic level of extranet is protected by L2 encryption and together with physical perimeter they compose domain that ensures information management for restricted level. This extranet is isolated from outer world and especially from Internet threats by Aircap  defence. There might be other domains at this same level of security but since their control is with other authorities they are divided from rest or basic level extranet by OSI level 3 encryption and firewalls. This is typical when for example Air Force wants to keep their domain perimeter intact but utilize Armed Forces general network services.

Within basic domain there is the extended level security for management of confidential information. COMSEC is using layer 3 encryption called Virtual Private Network, (VPN) thus disabling all attempts to tap communications from lower level. There might also be one way only -gateway between restricted and confidential levels of security. This “diode” is used to transfer data via transactions from lower level to upper level of security but not to the other way. There are many attempts to build these one way only -gateways but their integrity remains always relational similarly as it is with the protection from Aircap isolation. Where there is a human being as a link in this chain of security, he will remain the most vulnerable for exploitation by an outsider.

Functionally structured military organizations have been building these system, service or organization unit based domains while sustaining their control, authority and self-defined development. “Need to know” information management policy spread out from operational level down to tactical level. NATO and US military organizations faced results of this tendency back 2009 when then ISAF Commader General Stanley McChrystal, seeing his force ineffectively trying to share information via “sneaker network” (manual download, transfer, upload procedure), called for better information sharing to counter networked adversary. Afghanistan Mission Network program and quest towards “share to win” policy was created 2010 to integrate numerous separate domains within NATO and within National force structures as described in fig 7. 


Figure 7: Example of attempts to solve multidomain interoperability while maintaining domain trust based security

Since then a massive gateway structure together with aligned security policies have been built to allow different systems to be used via one core platform: NATO Secret Mission Network. National domains are attached to this core by Network Interconnect Points (NIP) that are cross domain gateways allowing users from NATO secret level to access any secret level system provided by contributing nations. This enable the shared use of about 165 applications of which 55 were critical to mission.  Mission Network was finalized when important part, Network Operation Centre (AMNOC), was opened to enforce common policy throughout federated domains. 

Because of tight schedule to build up this core network it was done with a cost of inner security. First there was no defence structure inside NATO secret core but later monitoring and reaction based security structure was adopted from US NIPR network.  This had started 2006 and was called Host Based Security System.

TRUST BASED ON MONITORING AND REACTION CAPABILITIES

As the number of federated domains and number of nodes interconnected increased, perimeter based security policy doomed to be futile. There was something to be done behind high walls of domain perimeters to control security and react when malevolent incidents were detected. This is again very basic military tactics to survey larger space of operation and react with tactically positioned rapid reaction forces to counter any insurgency. Host Based Security System , HBSS was created in US DISA together with McAfee and BAE Systems, who were given the contract to implement first monitor agents to all hosts in network. All hosts that either process data or store it were provided with agents or equal function to produce required surveillance information to Security Operation Centre (SOC). SOC has control over IT Operations change management and configuration management and can quickly change configuration when malicious behaviour is detected from information environment as illustrated in figure 8. 


Figure 8: An example of Host Based Security concept by HBSS and McAfee

Basic principle is to define baseline or ‘whiteline’ of existing normal features within IT-domain. Special detectors are created to detect anomalies out of baseline patterns. When anomaly is detected, alert is forwarded to SOC, which creates security incident and starts executing countermeasures. Another approach is more proactive where security engineers are using Sandbox environment and penetration tools to test vulnerability of each ICT-component. Components are patched and hardened before released to operations environment. This process when executed continuously keeps Operational ICT up-to-date and as sustainable as possible. There are always failures in systems that are not remedied yet. They are called 0-day vulnerabilities. To compensate these vulnerabilities, special thresholds around these components could be installed to detect any attempts of exploitation.

Further host based security requires additional information from possible “honey pots”, systems outside cross domain gateways, that lure possible penetrators to capture their attack profiles. There is also a need to connect with larger society of cyber experts that continuously monitor malevolent behaviour in networks, collaborate and make better sense of incidents in networks as depicted in fig 9.


Figure 9: An example of Security Operations Centre concepts within Host Based Security

Explained Deterministic and statistic approach may be complemented with more heuristic surveillance methods. They are capturing full packets from both core and front switches and using advanced big data analyses tools over a vast amount of data. This enables to capture different attack profiles and silent trends.  
 

NEW CHALLENGES OF ZUCKERBERGIAN ERA INFORMATION MANAGEMENT

From Gutenbergian culture of information management there is a long leap to Zuckerbergian culture of information management as illustrated in fig 10. The amount of information is increasing faster than ever before, produced by both people and machines online. Text is replaced by pictures and videos just because it is easier and people can capture essential information from those better than from plain text.


Figure 10: A point of view that illustrates the change in information management from Gutenbergian culture to postmodern culture that Zuckerberg’s Facebook era represents

Data and information policies of Gutenbergian culture cannot keep up with emerging magnitude and speed of Zuckerbergian era. Book takes too long to write and publish. Web page, blog or wiki are much faster and more able to manage larger amounts of information between people. Data virtualization will change information sharing between Things in Internet. Internet can transfer more data than anything this far. It can connect billions of both people and Things to create intertwined man-machine systems.

Information security based on sites, perimeters or hosts is not able to scale to Zurkerbergian era amounts of information, where Facebook is extending by 210 terabytes (1E12) per hour and 20 % of users per year.  Even in military domain amount of data is increasing exponentially with image capturing. General Gartwright estimated that when DoD is evolving to cloud computing and mission network of Things, a military force in one operation is managing about one exabyte (1E18) of unique data in a month.  As thousands of sensors create new data continuously publishing it in computing cloud where huge clusters of HADOOP  file system (Facebook’s biggest Hadoop cluster is handling about 100 petabytes of data ) store it and provide it to analysing tools. It is clear that file, host or relational database based security cannot provide required confidentiality, integrity or availability of information for this data revolution.

SERVICE BASED SECURITY TRUST

Last attempt to provide IT security for exponentially increasing data is to collect it as fast as possible from sources whether they are people or technical sensors and store it into logically centralized datawarehouses. Large amount of data is closed within inner structure of “cloud computing” platform which is better controlled and data integrity and availability is more easily sustained. Confidentiality is based on user identification and role based access via application or presentation layer. This security structure is based on onion model of so called “defence in depth”, where network, host, application and data based IT security measures create integrated solution as depicted in fig 11.


Figure 11: Some principles of onion structure of IT security

When network society grows too large to be cost-efficiently monitored and behavioural baseline changes too quickly to recognize any normality, one has to once again gather all information to ‘libraries’ and create strict access policy and librarians to enforce it. 

There is no trust on end user devices thus their integrity has to be restored by compliance check and reconfiguration. To do this effectively the baseline of end user device has to be simple and configuration “thin”. Terminal has come back to vocabulary from late mainframe and early Unix era. Terminal is very thin client that may only remotely operate any complex applications merely by keyboard, screen and mouse.

There is no trust on access network so all sessions has to be encrypted over untrusted and in many cases roaming access networks. There might be Mobile IP solution that enables terminal to roam between available and cost-effective network connections and most often encryption is integrated into Mobile IP application. 

There is no trust on user names or single passwords but stronger identification is utilized. This may vary from smart card based token that includes secret key to biometrics based authentication like fingerprints, face picture, hand vessel picture or retina. User and Service needs bilateral authentication to create a temporary trusted session over untrusted environment. This requires bilateral authentication protocol to verify both entities. Often third parties offer certification authority services to both instances.

Since there is no trust on incoming terminal nor the communications protocol it is using nor the content it is proposing, a complex access gateway between service base and outer world is needed. These gateways include number of firewalls, demilitarized zones and interruptive proxies with content checking and filtering.

Within “Computing cloud” there are several layers of defence starting from very first portal and access management layer, where users are allocated to roles and roles have authorized rights to run applications and process data. This enables to manage hundred thousands of users categorised into thousands of roles and further allow access to applications and data content. This is called identity and access management. All subjects are under constant surveillance to detect anomalies and kill intruders before they are able to fulfil their operation. There are different gateways for man-machine sessions and machine-machine sessions to detect more easily different behaviours.

Applications may have structured aligned with Service Oriented Architecture where each function is requested via Service Bus and applications themselves may locate in different instances. This requires particular application identification and authentication structure within SOA platform.

There are many service based security services available in modern time Internet. Most eGovernance/smartGovernance or online Financial services are constructed in this manner. In military domain there are similar features to be found in NATO Future Mission Network  and DISA All Partners Access Network . Examples of service based security implementations are illustrated in fig 12.


Figure 12: Examples of service based trust structures in IT security

CONTENT BASED SECURITY TRUST

If we are living in a world where platforms and connections are changing swiftly, there is no longer single layer of trusted platform to be defined. Besides people there will be billions of Things connected to Internet both producing and processing data. Definable server will evaporate somewhere in software defined network and cloud.  SOA and Semantic web will disperse monolithic applications to distributed applets that form temporary functions according to requests of each processes. Text files will change to structured data with both format that machines can understand and metadata that ontologies can make sense of. Databases will be gradually virtualised first as loose structure of several databases but eventually to a datafabric stretched over whole cyber space where data items are addressed as we are now addressing web pages.

There is a possibility to lay trust structure around all the data items if they are restructured in semantic web way and equipped with metadata. Another base of trust is the user identity, which is tied to data items via 3rd party certification authority services. This enables sensitive data to be stored, processed, presented and transferred within untrusted ICT-platform while optimizing its assurance according changing situations. This may be called also software defined security . See fig 13 for examples of transformation and semantic web security structure.


Figure 13: Evolution of Information Technology and examples of content based security structure

One possible way to provide new structure for trust in content based security is described fig. 14. Availability is provided by distribution of resources since there is no guarantee enough for one node or centre to be available as required. This is typical especially to military environment. All entities in space should have a unique address with which they can be found and addressed. Every entity has to be authenticated based on strong identity. Tokens and biometry are already there for people but strong machine identification and authentication remains to be implemented.

There has to be logical structure both in data and applications to ensure that they are what they seem to be and functioning as expected. Each data content should be encrypted or signed to ensure data confidentiality and authenticity. There is a need to implement separate certification authentication and key management structure which is one base of authority and identity.


Figure 14: Example of trust structure in content based information security

Encryption key management is following the chain of authorization from user to assigned application, which is provided with needed key material to be able to process user required data. 

Applications of content base security are still rare, but one can find them within some cloud computing environment and at very tactical Battle Management System level.

IT SECURITY ROADMAP FOR MILITARY ENVIRONMENT

This study started from very basis of our information security based on tangible container of information, a physical location of information and authorized process in accessing information. Based on these structures of trust, military have been building gradually their digitalized information management and accordingly IT security from pure site based approach to controlled domain and further trying to monitor and control all possible hosts that either process, transfer or store information. Now military is facing a major leap in magnitude, speed and format of information and they are advancing towards service based IT security and have visions of content based information security.

Since IT security is not an asset itself but always a quality of existing information structure where both people, processes, information and technology are changing there is no ultimate end state in this technical roadmap that is illustrated in fig 15. Since IT security is based on risk management, there are some categories of information that are using more technically advanced structures whereas top secret information still remains on paper and is stored in vaults. Military have been last 10-15 years following civilian IT security measures where finance and commerce have been leading implementers. 


Figure 15: A roadmap for IT security within military environment

On this map there are different roads that military may choose based on their risk management, organizational, process, information and technology maturity. There is the basic path of SITE-DOMAIN-HOST-SERVICE-CONTENT, which has been followed in NATO and DISA structures. There has also been examples of successful journeys through other paths. Netherlands Armed Forces showed example early 2000 when they were promoting TITAN –structure for mission IT structure. Parallel with this program Finland was piloting their Transferable Operation Centre, TOC concept. Both of them left domain based security and went directly to service based security.

MultiNational Experiment –exercise has been developing Service based structure since 2006  for Crises Management operations in order to have all stake holders collaborating with each other in complex crises situations. Finland supported MNE 4 exercise with Shared Information Framework & Technology (SHIFT) that “demonstrated method to share information across various military, governmental and non-governmental information domains”.  

Finnish Defence Forces took path from domain based to service based between 2007 – 2012 in their national operational command and control support. FinDEF did change from site based 22 domains to one private cloud computing based service structure where information services were provided from restricted to secret. It was not fully successful since some parts of Defence forces were not as mature as others so there was not equal starting point to this journey but implementation server as a model for others to follow.

DISA operated networks improved their perimeter defence with host based security early 2006.  Afghanistan Mission Network went through transformation from very domain based structure 2009 to more flat and host based security by 2013. 

Finnish Land Forces are on their way from site based security directly to content based security with their next generation Battle Management System and tactical communications M18. 

There are plethora of references I have not included in this revision. Please ask a copy with references.

2014-03-28

Lessons from developing Army Command, Control and Information System for Finnish Land Force during 2007 - 2009


This is a story of developing Army Command, Control and Information System, ACCIS for Finnish Land Forces. Development program was called MATI 2 (in Finnish MAavoimien TIetojarjestelma 2). This story is composed from personal observations of the author when he was the head of G6 and later inspector of signals between 2008 – 2012 in Army Command.


System engineering as development model

System engineering or V-model has been the main method in military systems development since 1950’s and famous especially from Apollo program. The model, which is depicted in fig 1, is originated from need to identify and manipulate the properties of a system as a whole rather than only properties of each part. This approach has been followed successfully to develop military systems, which has been integrated to force via manual processes. Problem has occurred when force integration requires machine to machine interoperability. Especially command, control and communications systems that has been creating a common bus for data exchange are facing development challenges. Defence vendors have legacy for closed architectures in the name of security, military coalitions are not flexible enough to define interoperability standards, fast evolving civilian information and communication technology is creating new pressures towards traditional military development and contemporary multinational operations have required quick solutions implemented during operation. Linear planning and investment driven development has been too slow to answer ongoing operations requirements.


Figure 1: Classical system engineering approach to development

This is a story of all good intentions to follow system engineering development model while mitigating risks by Commercial-Of-The-Shelf procurement, phased delivery via Initial and Full Operational Capability, special implementation of AQAP-quality measures and full involvement of functional, hierarchical organizations.

A short history of Army Command, Control and Information Systems in Finland


In 1994 there was the first field exercise in Finland, where main command post of an infantry brigade was connected by email service (VAXMAIL) to upper echelons. Common Operational Picture, COP was created still manually on paper maps and sheets of transparent plastic. Situational Information though travelled quickly by using digital messaging system, email or telefax. Orders were delivered verbally, written on paper and drawn on plastic sheet. Computers were used to help in creating documents, calculations and presentations within command post. Digital files were shared via Ethernet connections between offices, shared storages, closed email systems or by dispatchers with portable mediums.

In 1995 there was a field exercise where first Army Command, Control and Information System (ACCIS) was established and connected all the way from brigade command post to army corps and further to theater headquarters. It was called OPJO-95 (OPeratiivinen JOhtaminen), a Unix-based Graphical Information System (GIS) with virtual database to give processing speed, integrated message system to update COP -information and replication system that enabled transfer of larger amounts of data at time either on line or with courier. For first time Army orders were given on a digital map and delivered to subunits by electrical sheets over IP- or message networks. Typical Graphical User Interface (GUI) of Unix and even more complex GIS GUI made operator training time consuming and only very seasoned operators were able to beat “manual – pedal” planning and ordering system. This ended with situation where only those officers, that were forced to utilize system, possessed some skills to operate with it. Evolution of ACCIS in Finland is depicted in figure 2.

In 1997 artillery did use first time information system (JOHtamispaikkaLAitteisto, JOHLA-97) for their fire control centers. It was integrated with message system so forward observer (FOB) was able to send formatted message with messaging system -93 directly to JOHLA that presented fire command on digital map, calculated the position of FOB and delivered fire steering automatically to batteries. This really fastened indirect fires by removing human operators from calculation and signaling functions.


Figure 2: A short description of ACCIS evolution in Finland

The 2000 field exercise witnessed next generation ACCIS at every brigade level command post called JOTI JOhtamisen TIetojarjestelma). JOTI was a cousin of OPJO but with Microsoft GUI, easier graphical interface and it was running faster on a PC in each office of a command post. Staff work was accelerated by both office tools and GIS- system that enabled distributed working, cut briefing times half and squeezed planning time to one quarter from earlier. Since JOTI was integrated with message system all situational information was send in formatted messages to JOTI systems and semi-automatically presented in GIS-screens. Top – down development automated all COP –and Intelligence processes with brigade command posts but left lower echelons with manual maps and message terminals. Program failed to automate whole process from bottom up.

In 2005 JOTI update came up unsuccessful, sustenance of JOHLA-family ceased and Army Command had to make decision to downgrade their ACCIS to JOHLA-97, which had been updated without problems. Since functions were almost identical in both systems, there was no major issues in training. With sudden stop in their major ACCIS line (JOTI), Army Command was forced to initiate a program to procure and field next generation ACCIS called MATI 2 (MAavoimien TIetojarjestelma), since everyone understood that JOHLA-97 extension was only a temporary solution.

Initial phases of MATI 2 -development program


Army Command planned to have reasonably quick program to field next ACCIS called MATI. Since they did not have similar strong technical and tactical development force as was with OPJO/JOTI/JOHLA –family in early 90’s, Army Command decided to go after MOTS–product, which should have mitigated major risks of development. All national tactical requirements were changed to international since NATO –compatible was thought to be main driver at that time. National tactical symbols were changed to APP-6A and planning process was changed accordingly to NATO STANAG 2014. With Multinational Interoperability Program (MIP) there was also new datamodel for information exchange called C2IEDM (Command and Control Information Exhange DataModel). That was selected as a base for all datamodels in Defence Forces and Army Command required to follow model within the whole MATI-system.
Although MATI was planned to be bought as a turnkey solution mainly off the shelf, there was extensive planning and design work with field tests to capture primary requirements that all possible products would be tested against. Staff processes were changed to NATO compatible ones and they were role played and finalized as process documents to be utilize in further testing and training. Technical requirements were definite with Microsoft middleware platform, ability to operate standalone and within networked group over tactical non-IP communications. MIP –compatibility requirement was given from Joint –level and Service Oriented Architecture (SOA) was recommended although no specific requirements for open interfaces or structure were stated.

Possible vendors were screened with Request For Information –process that created lots of documentation produced from marketing point of view. Actual procurement process included Request for Proposals – Proposals – Prove of Concept – Shortlisting – Negotiation – Decision and contraction –phases. Prominent vendors were invited to intensive laboratory tests where both users and system administrators screened MOTS-applications. One vendor of Battle Management application (BMS) was chosen amongst many with good scores in usability tests and the only one that promised full MIP compliance with SOA structure. Major defence industry was either not able to provide functional system or they proposed old systems like existing JOHLA and JOTI.

Since chosen vendor was able to provide only BMS –application, a loose consortium was negotiated between application, communications and quality assurance providers. Joint level had created a theoretical model where 3rd party was utilized between customer and providers to help managing relations and deliverables. It remained unclear in contract structure, who was responsible of what deliverables and how they were approved. This model was new for all parties of consortium and customer. It was planned that application vendor provides off-the-shelf BMS-software to communications integrator, which in turn delivers BMS application with right communications configuration to customer. Because Defence Material Command had adopted AQAP -quality system, a 3rd party vendor was contracted only to do quality assurance.
Since this seemed straightforward linear procurement, customer distributed responsibilities in functional way as depicted in Fig 3. Material Command was executing procurement of Material parts supervised by G10 of Army Command. Army School was given testing and C3 processes development responsibilities. J6 did coordinate development with architecture, whereas C4I Centre did provide common components like Operation System, interfaces like MIP and XML gateways and joint level C3 system (LEIJONA). Defence University and R&D Centre were developing special functions supporting operational analyses. Everything was integrated by Inspector of Signals supported by G6 of Army Command, that also took care of other capability components of DOT(M)LPFII structure and prepared to field and operate the future system.


Figure 3: Distribution of responsibilities in MATI 2 procurement

Contract was signed for 2.5 years and deliveries were phased in iterative way: by the beginning of 2008 a integration version of BMS, by mid 2008 an Initial Operational version of both BMS and Communications and by the end of 2009 a Full Operational Capability versions to be fielded. Clear milestones and documented requirements to be fulfilled at time of delivery. This seems to be typical system engineering procurement and delivery project with responsibilities distributed to appropriate functions and risks mitigated by simple linear approach.

Challenges occur


Everything started well. All functions were in place and there was no big risk since BMS application was coming off-the-shelf. Contract was phased to three clear and measurable deliverables where all vendors had their own responsibilities. Funding was fastened for 3 years and risks were managed. Two important questions - WHY and HOW - were answered. But were they? 

Why this program was established and how shall it enable the capabilities of multiarms Battalion task force? Program was established to replace temporary BMS solution in Brigade command post and it was not planned to improve C3 processes or multiarms effect majorly. Sure all investment justifications were communicated within orientation for force digitalization and enabling awareness, but effectively there were no plans to extend BMS further than Battalion command posts, no plan how to integrate artillery, sappers, reconnaissance and surveillance or logistics processes into one system, no plan to integrate sensors or other awareness sources into system. From initial program intent project adapted only the simple task of replacing existing ACCIS.

How to integrate all layers together to create functioning system as defined in Fig 4? Only command post in Brigade had Local Area Network support. Battalions were still using analog radios and messaging system -93. Land forces was using their own messaging format that was not interpreted to MIP-datamodel. No one had an idea of how full synchronized BMS is able to replicate all necessary information throughout Brigade. There was no cross domain gateways (IEG) in between Land components and Joint C3 system nor were there any MIP-interoperable Joint C3 systems (LEIJONA) to be integrated to. No single model of the whole system either in small world installation or logical model was created to simulate and see how all these parts come together.


Figure 4: Technical challenges in ACCIS structure

Engineering logic was simple. If all these components were seen functioning separately, why should not they be easily integrated and function together. Unfortunately that was not the case!


Tactical communications challenges


As noted earlier, BMS was functioning over 100 M Ethernet between few nodes, but there was problem to fit replication traffic over existing Combat Radio Network. Communications integrator had assumed similar messaging as in older version of JOTI and JOHLA. BMS application session level required integrity of connection that messaging was not able to provide. Throughput of CRN digital channel was not near what was required by replication between BMS instances. Tactical IP connection with 128 kbps gross throughput was feasible but since Bit Error Rate was around 10 EXP-4 BER, the TCP-session was not stable enough for BMS client-server utilization. Application assumed fixed IP-addresses when mobility of command posts required changing addresses. Existing encryption was optimized for short messaging (32 bit block encryption) purposes and was not supporting continuous flow of longer IP-packets. 

At best tactical communications were useful if stations were fixed, running with maximum transmission and plain text. This was not what was required and communications integrator announced that project would need at least one year more and millions of euros additional resources.

Application developer was testing his system on LAN with very limited number of nodes. Communications integrator was honing existing messaging system with unrealistic assumptions. AQAP vendor did not understand this kind of system of systems development but concentrated merely in documentation transparency and governance.

Datamodel challenges


J6 had made architecture decision of adapting MIP information exchange model, C2IEDM and further JC3IEDM as a base for all Defence operational and tactical information management. They appreciated an idea of using one information model from soldier level up to Joint Command level since that would support best the transparency of common operational picture and provide interoperability for both national and international level. Decision was understandable from the view that MIP was the only information model development program running at that time, NATO indicated to adapt it fully and it was experimented successfully in exchanging COP information elements between nations CCIS entities.

Unfortunately no one had utilized model with larger force and with more exhausted replication. It appeared that application was not able to manage more than 2 000 to 10 000 units and thus was incapable to manage the whole Land Forces order of battle not to speak of Defence Forces. It also appeared that C2IEDM model implementation in relational database was using processing power extensively. Complex queries took long time to be executed within normal PC instance. C2IEDM replication protocol and policy was not planned to tactical communications thus introduced several errors and frequent disconnection from replication network. Replication was also meant to work between two instances which automatically restricted topology to a star. Even within Headquarters there was not possible to sustain longer than two level hierarchies in replication. Replication with three levels of hierarchy was not feasible.

C2IEDM as relational datamodel was not flexible for changes and did not understand recent features of operational space like NGO’s, supported population, communal administration, integrated effect of multiarms, jamming as fire element, etc. Datamodel took also long time to be established in first time. There was no practical way of cleaning old data but dumping whole system. If unit was not part of original Order Of Battle, ORBAT, it was impossible to add it on the run. All information was owned by troop that was part of hierarchical structure. Datamodel had problems to understand terrain, time or phenomenon based information. Datamodel was binding land forces tactics to international compromised understanding without national special features and did not support tactical adaptation or learning.

Testing challenges


There was no reference environment to test whole system of systems until it was installed for field exercise. This testing procedure did not detect main problems but until staff was employed in command posts and all vendors started parallel remedy onsite, which produced even bigger problems without shared knowledge of root causes. Everyone ended up frustrated and untrusting in each other’s capabilities and products. 

Several field tests were executed but not with same topology, ORBAT or situation, thus it was difficult to compare previous test results. There was also problem with staff officers using legacy procedures rather than NATO planning guide, GOP or Common Operational Picture fusing and sharing procedures. New SOP’s were not captured by testers, which produced one hand false expectations and other hand wrong perception of new system.

Basic performance and integration failures prevented test programs to proceed, which denied further feedback to developer. Challenges were managed more sequential than parallel, which did not give full picture of systems possibilities or challenges. Without laboratory environment and reference system there were no tools for gradual/parallel testing nor bringing in testing crew as technical issues matured. Lots of resources were lost in full scale field tests and MATI 2 got very negative brand value, which was further communicated by testers coming from different units.

Management challenges


Managing whole DOTMLPFII capability development was challenging since main focus was on material component. There was no aligned or coordinated plan to proceed with other components. G6 tried to come up with holistic plan, but major parts were missing. Army School did good job with staff work and BMS coordination, but Signal School was outside with their own agendas. A value chain of MATI 2 development and support organization is depicted in Fig 5.

Doctrine connection to tactical or C3 process development was nonexistent. Other arms were missed and most importantly intelligence, reconnaissance and surveillance was not included. The big Why was lost since initial commander’s intention was lost during haste execution and along many partially optimizing tentacles of functional organization.

Command post organization and C3 procedures were not really questioned although G6 tried to define major processes and published Land Forces C3 doctrine. There was a major mistake in attempt to adapt NATO based procedures and terminology as part of national art of operation and tactics. Outcome was long unsureness of used terminology and basic procedures.

Training focused only to have field exercise testers understanding of new processes and skills. Further fielding was not planned and education of new basic skills remained to be initiated. One should start to change basic tactical understanding amongst leaders for them to first understand WHY and then be able to instruct the HOW to troops when skills based training starts.

Even in material issues there were challenges since Material Command was not able to provide terminal devices for field use. C4I Agency failed to provide hardened operating system to field terminals. Since Joint operational C3 program failed totally, land forces were left alone without provincial above level C4-system to connect with and without necessary information services. Information Assurance solutions that C4I Agency proposed were not applicable in tactical environment. Even with best technical solutions off-the-shelf, there is no capability if understanding, integration and training skills of the whole system of systems are missing.

Leadership and management doctrine was still based on authoritarian and deep line organization without idea of emphasizing mission command or self-synchronized troops that would be able to swarm around adversary using terrain like infantry used to do during II WW. It was remarkable that conscripts and reservists adapted new tools and methods faster than enlisted, NCO’s and officers. It seems that learning new things and adapting new habits requires first unlearning of old things which takes both time and effort.

There was haste to build up required service support maturity within Signals personnel, since both technology, interconnections and system of systems interdependency required more Information Technology System Management, ITSM process approach than hierarchical line organization orders and control. If there is no legacy for process driven value chain like in artillery or logistics, then simple provider-client cooperation might appear hard to adapt by line commanded troops and their authoritarian leaders.

Maintenance and repair personnel were mainly sustaining individual devices that did not need software updates or complex configurations. Complex networked software defined systems required online warehousing, were all devices were powered and connected to maintenance network. Whole fleet of systems required weekly updates to keep interoperability and minimize vulnerability.

Information management was started to create and maintain training ORBAT for blue, neutral and red parties. Since there was a major delay at joint level C3 systems development, land forces were required to create alternative information feeds that provided weather forecast information, interagency Blue Force Track, BFT information, enemy target information from air and sea, logistics information from Defence Forces Enterprise Resource Management system, etc.

Interoperability required a standing Battalion task force reference system of fighting systems to be implemented. This reference was a small world where different experts and vendors met in integration testing, experimenting, system hardening, change management, root cause analyses and configuration management. The Reference environment sustained the integrity of complex interconnected system of systems and enabled quick changes when required.

Program management did have long hierarchical and administrative heavy lines of command that were slow, loosed information, prevented to create holistic awareness and were prone to small function optimization. Neither hierarchy nor customer driven management was followed but each function acted individually as they seemed best without understanding of causality and when failed they were protected by thick layer of irresponsible administration.


Figure 5: Planned value chain of MATI 2 development and support organization

The value chain pictured in Fig 5 never functioned since there was no maturity for matrix cooperation, lines of communication were long, functional priorities of each sub provider caused friction and holistic understanding of chains operation was missing.

Getting out


When IOC tests failed a third time in a row, it was clear that this program has failed as a whole. Both communications integrator and quality vendor had pulled out in silence and only customer and BMS provider were left in confrontation situation. When analyzed the procurement process it became clear that there was no clear violation of agreement since both sides had misunderstood intentions and failed in communicating requirements.

It was responsible single people that were managing this situation since organizations were busy doing their administration. Inspector of Signals, head of G6, project manager from G10 and project engineer from Material Command together with BMS vendor came up with settlement which they proposed to Ministry of Defence where the final decision was made. 
Settlement included dismissal of original agreement by both sides. It included a new contract to develop BMS solution to a smaller international force and sustain it to the future.

Success in the end


G6 of Army Command took responsibility to manage this new international capability. They flattened development organization in to Integrated Project Team where representatives from G6, G10, Material Command and vendor worked together. It also helped that vendor had recruited two experts from Defence Forces to help in delivering understanding trough entire program. Since Special Operations Regiment had best experience in using BMS from patrol level upwards, they were tasked to test and develop holistic capability. Signals School was included with task to build IP-based mobile communications infrastructure for peace keeping operations for a company sized force. Development plan with parallel paths of advance was created, clear intermediate goals set and effort to communicate these goals to all stake holders was intensified.

As result BMS was fielded to international operations within companies that were operating in Afghanistan and in Lebanon. Both systems were integrated with other nations either at BFT or MIP –levels to create situational awareness over all forces committed within the same area of operation.

Lessons Identified


There rarely is a military capability that one can purchase as-is from vendors. Capabilities are often composed from DOTMLPFII parts and material component is nothing without other components adaptation.

As fighting system of systems becomes more complex and interrelated there is rarely a one vendor that is able to provide even all technical parts. The basic system engineering model is not best practice in developing complex system of systems capabilities.

Organization is often slave of its origins, culture and legacy systems. If one is not careful, future development is only renewing existing capabilities and not seeking asymmetric and strategic advantages over possible opponents. Since partial optimization and egocentricity is a risk for all projects a program or portfolio level should be very careful to overcome these pits. 

There are cases where lack of maintenance or negligence in updating key modules of one system of systems has led to dispersion of the whole capability. Software defined and interconnected systems should be kept up to date or their will easily flip from strength to weakness.

If single arm is concentrated only to improve their inner capability, there is a danger that it would not support multiarms operations. With integrated force there should not be any separate development program nor project but at least coordination with clear milestones at force capability level. NATO and US Forces have utilized ongoing operation in Afghanistan to improve their integrated effects.

System of systems capability is not based on single subsystem functions but subsystems cooperation. Planning, design and testing should include always whole system of systems structure – either live or simulated.

All datatransfer issues in tactical environment should be modelled and simulated to have better understanding of volatile environment and evolving requirements. If tested scenario is following only existing procedures or topology, there is a danger of major blindness.

If no other modelling is utilized then a small world of larger system of systems implementation should be created as reference environment. Around this sandbox one can both test all changes in configuration but also experiment with possible new components.

One should be careful in introducing long term architecture decisions since installed legacy does not follow new good principles, but needs to be integrated. When introducing datamodel practices one should have clear understanding of the magnitude of one’s force and sample data from existing and possible situations. With this understanding one should start vigorous modelling and simulation before advancing to any implementation. Laboratory environment does not full truth and field tests are very expensive way to find out design failures.

Data and information flow in tactical domain is very much dependent on tactics and battle technics used. Adopting models from foreign forces does not give similar capability to your force until you change whole stack of DOTMLPFII components.

If customer is not able to create testing procedures or test environment with use cases, there is no clear understanding of required capability and the way how new subsystems are integrated to existing structure and processes. Testing, even in systems engineering way, is a sequence of tests that should utilize information collected from whole test track record. 

If so called ‘show stopper’ emerges during testing, one should have parallel courses of action to proceed. Otherwise one failure in pure sequential procedure will become a major risk for whole project. One should not neglect failures but encapsulate them in order to proceed with other features and come back next time with remedy.

Modern software defined processes may be developed either tailored to customer defined processes or adapted as-is implemented in turn-key system. If force is not aiming to follow fully for example NATO based art of operations, tactics, order of battle and has similar weapon systems, one should not adapt NATO based processes, symbols or terminology as-is. Adaptation will most probably fail since there is no required maturity of tactical understanding or battle technical skills. Old habits are hard to get rid of at staff processes and combat level leading. It requires multitude of simulation training to get rid of old habits and adapt new ones.
With software defined technical systems also maintenance and sustenance are to be changed from legacy model of ‘box’ or component based repair. Immaterial asset management and software updates become major challenges for logistics.

Document driven development is suitable for small system development and quick delivery projects. ‘Sprints’ should have clear stories and prompt delivery. When complexity and interrelationships increases a more iterative development approach should be implemented with several parallel courses of action. Spiral development with many practical testing phases through laboratory, reference and pilot environments is more risk driven approach that is recommended for CIS system of systems development programs.

Normal military planning with different scenarios, decisive points, commander’s intention and ‘WHAT IF’ sensitivity testing should be implemented also in capability development programs. Clear customer driven approach normally helps to align number of different units along value chain because people has tendency to do meaningful work.