The automatic detection of attacks and threats is a challenging task. To tackle this problem, we develop self-learning methods that adapt to changing conditions and enable inferring patterns of malicious activity automatically. This research ranges from network intrusion detection to mobile threat analysis.
We are interesting in analyzing and understanding the security of machine learning. We aim at constructing learning-based systems that are resilient to different forms of adversarial inputs, such as evasion and poisoning attacks. Our work ranges from robust malware detection to general concepts of adversarial learning.
Analyzing and understanding malicious software is a time-consuming task. We develop methods for automatically analyzing the structure and behavior of code. Our methods allow for efficiently screening and examining large amounts of malicious code, such that new threats can be better identified and contained.
We develop method for discovering and analyzing vulnerabilites in computer systems. For this research, we blend methods of code analysis with machine learning. This combination enables us to create intelligent methods that can locate security flaws and privacy issues in software with little human intervention.