Output details

Can a good offense be a good defense? vulnerability testing of anomaly detectors through an artificial arms race

<19> Significance: Published in Applied Soft Computing, a prestigious Evolutionary Computation journal. Originality: This work takes a novel approach to intrusion detection by utilizing an arms race between an artificial attacker and a detector. Rigour: While attacker attempts to evade detection by producing stealthy variants, detector in turn adapts its detection model to detect multiple variants. Vulnerability testing is automated to be more proactive in eliminating detector vulnerabilities and increase detection coverage. Kayacik conducted all the experiments and wrote 70-80% of the content.