PhD-Improving the efficiency of AI models for malware analysis F/H

about the role

Your role is to carry out thesis work on: Improving the efficiency, interpretation and robustness of AI models for malware analysis” 

· Global context and problematic of the subject

Use our expertise in cybersecurity to combat future malware more effectively, as it becomes increasingly powerful and complex.

AI models for the detection of cyber threats are coming up against the ever more rapid development and refinement of attackers, now aided by generative AI, and so-called Malware-As-A-Service markets.

Through new AI models, improving robustness and explainability will help to fill these gaps, relying not just on data, but also on business knowledge.

· Scientific objective - results and challenges

The aim of the thesis is to propose a generic framework for integrating business knowledge and expertise, in order to improve AI models for cyber threat analysis, with a use case in malware detection.

Main challenges:

1) Mastering and integrating new AI technologies (LLM, neuro-symbolic AI, etc.) to address cyber issues.

2) Improving collaboration between cyber players (analysts, reversers) and AI/ML researchers.

3) Integrate business knowledge on malware (YARA, Mitre Att&ck, empirical knowledge...).

4) Address issues related to the robustness and explicability of models and improve confidence.

Main expected results:

1) State of the art on neuro-symbolic AI and LLMs, and their combined integration with analysis models for cybersecurity.

2) Proposal of a generic framework applicable to malware detection, and to other cybersecurity topics.

3) Implementation of a prototype for advanced malware detection and analysis