27.06.2013 Views

6th European Conference - Academic Conferences

6th European Conference - Academic Conferences

6th European Conference - Academic Conferences

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

2. Network infiltration detection<br />

David Merritt and Barry Mullins<br />

Intrusion detection helps us answer the question: “Is there a malicious intrusion into the network?”<br />

Because there are countless manual and automated mechanisms to identify suspicious network<br />

behavior, this section will only discuss the most common techniques for intrusion detection. This<br />

glimpse into intrusion detection serves as a backdrop for the explanation of the synthesis approach,<br />

which assumes that network infiltration can be detected somewhat reliably.<br />

A network-based intrusion detection system (NIDS) detects network-oriented attacks and traditionally<br />

monitors the access points into a network. If a cyber spy chooses a common network attack method<br />

to infiltrate a network, such as a common buffer overflow exploit, then the NIDS will have a high<br />

detection success rate (Patcha and Park, 2007: 3448-3470). If there is a novel or sophisticated attack<br />

that is difficult to detect, NIDS relies on its anomaly detection capability. Kuang and Zulkernine (2008:<br />

921-926) have shown that an anomaly-based NIDS employing the Combined Strangeness and<br />

Isolation measure K-Nearest Neighbors algorithm can accurately identify novel attacks at a detection<br />

rate of 94.6%, where the detection rate is defined as the ratio of correctly classified network intrusion<br />

samples to the total number of samples.<br />

3. Malware detection<br />

Malware detection helps us answer the question: “Is there something malicious happening on a<br />

host?” This section is not an exhaustive survey of all malware detection mechanisms and methods.<br />

Rather, it simply makes evident the fact that there are numerous ways to reliably detect most malware<br />

on a system. Malware comes in many forms with many names. For simplicity and convenience, we<br />

will refer to any unwanted and malicious program or code running on a system as malware. Naturally,<br />

detection of unknown malware is the goal, assuming the cyber spy will use sophisticated, novel<br />

malicious programs to establish footholds on a computer and within a network.<br />

3.1 Antivirus<br />

Antivirus, or anti-malware, software does not need much explanation as it is a commonly used and<br />

moderately understood term. Antivirus products rely primarily on signature-based detection, although<br />

most products have integrated at least a rudimentary mechanism for behavioral analysis of<br />

executables. The vast majority of known malware is caught by commodity software. As a point of<br />

reference, most antivirus products have proven they can detect malware in sample sizes of over one<br />

million with accuracy in the upper 90 th percentile (Virus Bulletin, 2008).<br />

3.2 Malware analysis<br />

There are historically two methods of analyzing unknown programs, or binaries: static and dynamic<br />

(Ding et al, 2009: 72-77). Static analysis starts with the conversion of a program from its binary<br />

representation to a more symbolic, human-readable version of assembly code instructions. This<br />

disassembly ideally takes into account all possible code execution paths of the unknown program,<br />

which provides a reverse engineer with the complete set of program instructions and therefore inner<br />

workings of the unknown program’s code. Analyzing this code to discover a program’s purpose and<br />

capabilities makes up the bulk of static analysis. Christodorescu et al (2005: 32-46) and Kruegel,<br />

Robertson and Vigna (2004: 91-100) discuss a couple effective approaches in using this kind of<br />

analysis to detect and classify unknown malware.<br />

On the other hand, analyzing the code during execution is called dynamic analysis. Dynamic analysis<br />

is effective against binaries that obfuscate themselves or are self-modifying. This is due to the fact<br />

that the destiny of all programs is to be run on a system, so when the program is running, its behavior<br />

and subsequent system modifications can be seen. Willems, Holz and Freiling (2007: 32-39) and<br />

Bayer et al (2006: 67-77) discuss dynamic analysis techniques that are successful in detecting<br />

unknown malware. Also, Rieck et al (2008: 108-125) used a learning based approach to automatically<br />

classify 70% of over 3,000 previously undetected malware binaries.<br />

4. Data exfiltration detection<br />

Data exfiltration detection helps us answer the question: “Is someone stealing data off the network?”<br />

Detecting suspicious and outright malicious events in the realm of data exfiltration is arguably the<br />

most difficult but most important to achieve out of the three steps of cyber espionage. Because the<br />

existence of a computer network implies the need for data to be accessed both inbound to and<br />

181

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

Saved successfully!

Ooh no, something went wrong!