Maximize
Bookmark

VX Heavens

Library Collection Sources Engines Constructors Simulators Utilities Links Forum

Dynamic detection and classification of computer viruses using general behaviour patterns

Baudouin Le Charlier, Morton Swimmer, Abdelaziz Mounji
July 1995

1
[Back to index] [Comments (0)]

Abstract

The number of files that need processing by the virus labs is growing nearly exponentially. Even though only a small proportion of these files contain new viruses, each file requires examination. The normal method for dealing with these files in the virus labs is still brute force manual analysis. A virus expert runs several tests on a given file and delivers a verdict on whether it is virulent or not. If it is a new virus, it will be necassary to detect it. Some tools have been developed speed up this process. These range from programs that identify previously classified files to programs that generate detection data. Some antiviruses have built in mechanisms based on heuristics that enable the antivirus to detect unknown viruses. Unfortunately all these tools have limitations. In this paper, we will demonstrate how an emulator is used to monitor system activity of a virtual PC, and how the expert system ASAX is used to analyse the stream of data the emulator produced. We use general rules to generically detect real viruses reliably, and specific rules to extract details of their behaviour. The resulting system is called VIDES and is a prototype for an automatic analysis system for computer viruses and possibly a prototype anti virus for the emerging 32 bit PC operating systems.

[Read the article]

deenesitfrplruua