Heuristics - Understanding the static and run-time behavior
A milder form of hueristics called passive hueristics is employed by anti-virus companies where they parse the code statically and apply heuristics to deduce the potential of a virus. One such technique is the use of Markov Chaining and variants of Markov Chaining, which probabilistically tells if a given piece of altered code might bear the resemblance of a known family of metamorphic virus.
Heuristics is the anti-virus community's answer to metamorphic viruses which are the most potent class of virus to exist to date. However, as I argued in Part 1, the effectiveness of using heuristics is questionable. Many authors have argued that only 30% of metamorphic virus forms can be identified by this technique. Another side-effect of this technique is the raise of false-positives.
Behavior Blocking/Intrusion Detection (IDS)
Rather than wait for a new virus to emerge, propagate, destroy and then come up with a solution, intrusion detection technology shields a host (be it a computer or an entire network) and actively monitors every single operation. When a virus is about to unleash itself, these intrusion detection systems throttle and quarantine them, thus effectively diffusing the problem before they occur.
To understand this better, the following illustration shows the difference between how an Anti-virus technology and an Intrusion Detection technology handles a potential security problem. Imagine a malicious person trying to rob a local bank. In most situations the local bank would report the incident after it has occurred to the police who would then study the forensic, go after the clues the robber had left behind and then nab him. This is exactly how Anti-Virus technologies work.
Think of an alternative way. Imagine commissioning a secuirty-guard whose keen eyes are trained to capture the thief even before he has performed the action. For example, such a guard could look for patterns such as, “Does any person wear a mask? Does he appear to be loaded with a rifle? Does this person appear to be in a hurry?” etc. The idea here is that to a well trained eye, such activities raise alarms which can prevent the robbery even before it has occurred. This is what Intrusion Detection systems do in principle.
Two subclasses of Intrusion Detection are available - Host Intrusion Prevention Systems (HIPS) and Network based Intrusion Prevention. Since, this article is about Desktop Security, I would like to limit the discussion to HIPS. The biggest problem with such Intrusion Detection system is that they raise lot of false alarms. The vexing question is how do you guarantee that all genuine transactions are allowed? The problem with Intrusion Detectors is that they block several genuine operations and ask for validation. This makes these systems less reliable, noisy and cumbersome to use. One solution to alleviate this problem is the White Listing technology which is explained later.
Technical Note: The term Behavior Blocking and/or System Firewalls are equally applied to HIPS technologies.
Threat Scoring
Black Listing and White Listing Techniques
Anti-virus products maintain a list of malicious programs to compare against. They maintain either the signature of the malware or the the hash (MD5, SHA256) of the actual malware. This list has come to be known as the Black List.
Consequently, corporate IT security managers should focus on allowing only the known good programs rather than focusing on preventing the bad ones. This is a fundamentally different approach and requires a very different mind-shift. The technology behind it is called the white listing technology which is the exact opposite of black list. i.e the signature and/or the hash (MD5, SHA256) of the good programs are maintained in a list. During execution time, only those programs that are part of the white-list will be allowed for execution.
Managing the Gray List and the Power of Peer Review
One of the popular techniques is the use of social networking. For example, peer-reviews are one option by which the classification of gray programs can be administered. If more users rely on an "unknown program", then it is perhaps benign to allow such programs for execution. This is exactly the same approach that popular websites like Google, Wikipedia etc take. While one might argue the effectiveness of peer-reviews and social networking, collectively, there is no denial that such techniques will work. According to Surowiecki (2004), the decision making of a group is often better than that of individual users. Thus, often times managing the gray list via peer-review is better than asking the user for a decision.
Firewalls
Note: Sometimes the term Personal Firewalls are applied to that class of network firewalls that are installed on host (personal) machines. ZoneAlarm (Free Edition) is an example of a personal firewall that protects the host from attacks from outside entities.
Sandboxing/Virtualization
Therefore, in theory, the potential for damage done by any application is limited to the virtual container alone- which in most cases can be rebuilt with ease. Futhermore, sanboxing is somewhat similar to behavior blocking in the sense that all sanboxes will have to monitor for behavior at run-time and enforce policies at run-time. However, unlike behavior blockers, sandboxes limit the exposure to a virtual container. This introduces additional layering of resources (memory, filesystem etc). This tend to cause process execution delays.
Note: The term virtualization is often used with sanboxing technologies. Several emulators (hardware, OS etc) fall under this category.
Anti-Virus Technology Comparison Chart (Scroll below):
Action | Strength | Weakness | Scope of Monitoring | Intrusiveness | Examples | |
Signature Based Anti-Virus | Passive | Omnipresent, Effective against static malwares | Ineffective against polymorphic and metamorphic viruses | Files, Memory | Low | Symantec, AVG, McAfee, Kaspersky etc |
Heuristics | Proactive | Can detect known family of viruses | Moderately successful against unknown family of viruses | Processes | Low | Symantec, AVG, McAfee, Kaspersky etc |
Behavior Blocking (IDS)/HIPS/System Firewalls | Proactive | Can detect all intrusive behavior |
Cannot detect static viruses | Processes- Full lifecycle | High | DriveSentry, PervX, FireStorm, WinPooch |
IDS With Threat Level Scoring | Proactive | Aids in user decision making | Moderately noisy | Processes- Full lifecycle | Moderate | DriveSentry |
IDS with Threshold Based Threat Level Scoring | Highly Proactive | Very Low Intrusiveness | Processes- Full lifecycle | Low | DriveSentry | |
Process Firewalls | Proactive | Monitors processes at startup | Not granular enough | Process startup time | High | |
Personal Firewall | Somewhat Proactive | Monitors network processes and traffic | Does not cover all processes | Network Processes Startup time | Moderate | ZoneAlarm, Norton Personal Firewall |
Anti-Spyware | Proactive | Effective against spywares (tracking cookie etc) | Do not supplement traditional anti-viruses | Mainly web (browser) based processes | Low | Microsoft Anti-Spyware, ePestControl,AVG |
White Listing | Proactive | Improves system integrity | False positives | Files | Moderate | Bit9, DriveSentry |
Gray Listing (With Peer Review) | Proactive | Improved decision making | Reliance on external user’s knowledge | Files | Low | DriveSentry |
Sandboxing/Virtualization | Passive | Robust security | Slows down system performance, additional licensing cost | Virtual Memory | Low | VMWare |
