Malware Detection via Graph Analysis
Malware detection is a challenging task aimed at discriminating and recognizing malicious applications. This task requires a deep understanding of the static and dynamic properties of a program, and it is often carried out manually by experts. The use of human expertise represents a critical bottleneck in performing a timely analysis of a number of applications of increasing complexity. A possible solution is offered by machine learning tools that, relying on data, may assist, or even substitute, experts. Several machine learning approaches have been applied and tested on malware data; a particularly promising approach that is the object of much research is based on the definition and automatic processing of graphs.
The aim of this project is to study, evaluate and apply graph processing algorithms to the problem of malware detection. This would require developing a solid understanding of the available graph analysis methodologies; to assess their potential and their limitations; and apply these techniques to simulated or real-world malware data.