A Survey on Algorithmic Techniques for Malware Detection

Abstract

Malware is a specific type of software intended to breed damages ranging from computer systems fallout to deprivation of data integrity and confidentiality. Recently, along with the high usage of distributed systems and the increasing speed in telecommunications, the early detection of malware constitutes one of the major concerns in information society. A strong advantage that malware employs in order to elude detection is the ability of polymorphism (metamorphic or polymorphic engines). In this work we present efficient algorithmic techniques that, leveraging higher level abstractions of malware structure, perform an isomorphism check in malware's produced graph structures, such as function call-graphs and control flow-graphs, in order to detect every possible polymorphic version of a malware. Moreover, we propose an algorithmic approach for malware detection which focuses on the use of behavioural graphs as a more flexible representation of malware's functionality with respect to its interaction with the operating system. The main idea of our approach is mainly based on behavioural graph similarity issues.