Metamorphic virus detection based on Bayesian network

Metamorphic virus detection is one of the most challenging tasks of antivirus software and the most difficult ones are among known viruses. In this article we have used Bayesian network to recognize these kinds of viruses. The body of these virusesis made of assembly codes. At first opcodes are extracted as 1-gram from virus body, these opcodes are known as the characteristics of Bayesian network extracting these characteristics reduce dramatically the Computational complexity, memory and time used. After that, it´s time to draw the Bayesian network before drawing, Bayesian network should be training. Bayesian network learning is known as a NP-hard problem because of this utilizing exploratory research has proven that it can be helpful in a lot of cases; in which we have used hill climbing algorithm. This method is compared to different Hidden Markov Model and the methods of role-opcode are also compared. Experimental result shows that, utilizing Bayesian networkthe accuracy of virus detection increase, and other classify are not superlative that Bayesian network.