Cyber Thursdays Conclude on an Inspiring Note!

The 2024-2025 season of Cyber Thursdays came to a close on 6 February, wrapping up with an afternoon filled with insightful discussions and reflections.

At the intersection of cybersecurity and artificial intelligence, this Cyber Thursday took the form of a series of conferences where speakers from both industry and academia shared their thoughts and innovations.

Protection Against AI Model Theft

Marie Paindavoine, PhD in cryptography and founder of Skyld AI, kicked off the session with her presentation, Model Extraction: Why You Should Care and How to Protect Yourself?. She shed light on the risks associated with AI model theft in unsecured environments, explaining the consequences of such attacks and introducing reverse engineering and protection techniques.

Differential Privacy and Data Protection

Elie Chedemail, postdoctoral researcher at the Cyber CNI Chair of IMT Atlantique, followed with Differential Privacy and Graph Signal Denoising. His talk provided a clearer understanding of differential privacy and its role in protecting sensitive data, illustrated through a denoising method applied to mobility data.

Detecting Malicious Scripts with AI

Pierre-Adrien Fons, R&D Engineer, and Sylvio Hoarau, CTI Malware Analyst at GLIMPS, continued with PowersheLLM: A Large Language Model Facing the Horror. They discussed the challenges of detecting PowerShell scripts using an LLM trained for code completion and characterisation, focusing on model adaptation, achieved results, and errors identified within training datasets.

Security of AI Systems

Anne-Laure Wozniak, research engineer at Kereval, then took the stage with a presentation titled Testing the Security of AI Systems. She highlighted vulnerabilities specific to language models and the requirements imposed by the European AI Act, illustrated through testing on a generative AI system applied to children’s story creation.

AI and Disinformation: Geopolitical Challenges

Finally, Hugo Loiseau, Full Professor at the École de Politique Appliquée at the University of Sherbrooke, concluded the afternoon with a talk on The Geopolitics of Cybersecurity: Societal Impacts of AI on Disinformation. He analysed the growing influence of generative AI on disinformation, highlighting the challenges in combating these emerging threats and their repercussions on society and democracy.

See You in 2025 for a New Season!

This edition marks the end of an exciting season, underscoring the increasing importance of cybersecurity issues. A big thank you to all speakers and participants for their commitment and the quality of the discussions. This Cyber Thursday, held exceptionally in the afternoon as part of the Winter Research School programme, closes the season on an inspiring note.

New perspectives are emerging for the next editions, and we look forward to seeing you in October 2025 for an even more enriching second season!