27.06.2013 Views

6th European Conference - Academic Conferences

6th European Conference - Academic Conferences

6th European Conference - Academic Conferences

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

David Merritt and Barry Mullins<br />

to the probability of cyber espionage, it is possible to use this score to measure the probability of the<br />

entire event as being cyber espionage.<br />

It is important to note that an individual event detected by itself may not express outright if a<br />

circumstance is cyber espionage-related or not. A targeted, socially-engineered intrusion might be a<br />

sophisticated spam or phishing attempt. New and undetected information-stealing malware could be a<br />

new variant of benign adware. A consistent transfer of significant amounts of encrypted data could<br />

end up being an authorized VPN (virtual private network) connection. The subtlety of a covert channel<br />

may be difficult to detect or declare with certainty its intentions, but it does serve as an impetus to<br />

investigate further to determine the context of the channel.<br />

The advantage behind this synthesis approach is that intrusion, malware, or exfiltration detection can<br />

be viewed within the context of the whole event. Not doing so could lead to incorrect conclusions<br />

being drawn from insufficient context. But each step of a cyber spy's attack methodology is not of<br />

equal value to the investigator. For instance, it is a challenge to judge the intent of malware simply by<br />

looking at its detectability and functionality. Many malicious programs have the same functionality but<br />

are used for different purposes. In fact, many legitimate programs are frequently used maliciously<br />

(e.g., Remote Administration Tools). On the contrary, a targeted intrusion that is industry-relevant<br />

hints at the intentions of the adversary—to quietly get to specific targets. Thus, the intrusion factor<br />

should be weighted more than the malware “factor”. Similarly, with data exfiltration detection,<br />

covertness of the channel and sensitivity of the data are significant factors affecting the<br />

characterization of espionage. These factors' weights should have more weight than the malware<br />

installation factor.<br />

This strategic weighting of indicators is integrated to establish the Espionage Probability Matrix (EPM)<br />

framework, shown in Table 1. The EPM is used to determine an EPM score based on varying degrees<br />

of espionage probability, as indicated by the three columns of High, Medium, and Low Probability.<br />

Each indicator is assigned a value associated with its column, with High, Medium, and Low indicators<br />

being assigned values of 3, 2, and 1, respectively. The values for the indicators (e.g., targeted and<br />

tailored network intrusion) within each factor (e.g., Intrusion) are averaged to provide an EPM score<br />

for that factor. For example, a network intrusion that is not targeted at a specific user/group but<br />

contains somewhat tailored content results in an Intrusion factor EPM score of 1.5. This is calculated<br />

by averaging the “Not targeted” indicator (i.e., 1) with the “Potentially tailored” indicator (i.e., 2).<br />

Table 1: Espionage Probability Matrix (EPM)<br />

Intrusion<br />

Malware<br />

Exfiltration<br />

High Probability Medium Probability Low Probability<br />

Targeted; specific victims<br />

Tailored; social engineering<br />

required<br />

Novel or unknown<br />

Advanced info/data- stealer<br />

Covert channel<br />

Custom encryption<br />

Significant data transfer<br />

Industry-specific<br />

Potentially targeted<br />

Potentially tailored; social<br />

engineering may be used<br />

Not well known; variant of<br />

known<br />

Info/data- stealer<br />

Attempts to hide channel<br />

Standard encryption<br />

Non-trivial data transfer<br />

Partially industry-specific<br />

Not targeted<br />

Well-known methods<br />

Well known<br />

Not info/data-stealer<br />

No attempt to hide channel<br />

Not encrypted<br />

Negligible data transfer<br />

Not industry-specific<br />

The Intrusion row has a α multiplier, where α > 1, to represent the relative importance of intrusion<br />

classification to the overall EPM score. The Exfiltration row has a β multiplier, where β > 1, to<br />

represent the relative importance of data exfiltration classification to the overall EPM score. This<br />

effectively assigns greater importance to the factors that deserve it, as discussed. These individual<br />

probabilities are brought into context of the entire event by calculating an overall EPM score using the<br />

following equation:<br />

EPM Score = α·Intrusion + Malware + β·Exfiltration<br />

184

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!