DOCTORAL CANDIDATE
Bhavani Sunila

SUPERVISORS
Françoise Sailhan
Olaf Maennel
Kaie Maennel

Computer science, cyber security, artificial intelligence, machine learning

Anomaly and intrusion detection systems have been improved by recent advancements in artificial intelligence and sensor technologies. Yet these systems are still at risk from the evolving nature of threats, especially adversarial AI attacks. This issue is one of the main catalysts for this research proposal, which focuses on the development of novel AI-driven algorithms which may be employed to detect adversarial data poisoning in heterogeneous log formats. Methods for human-in the-loop models can be used to incorporate humans as expert knowledge to obtain better accuracy and trustworthiness in their decisions. Key parts of the study shall be log parsing, evasion attack detection, real-time analysis, scalability, and provide standardized benchmarks for evaluating these systems.