Is Machine Learning Security IFS-business as usual?
Teddy Furon (Inria, Rennes, France) – On-site presentation
The talk conveys a vision of Machine Learning Security based on Information Forensics and Security. In a nutshell, the IFS community protects three cardinal values (confidentiality, integrity, and property) of informative content, be it a transmitted signal (physical layer), an image (watermarking) or some text (fake news detection) for instance. To what extend can we cast this framework onto Machine Learning? We will see that it makes sense for some scenarios (where we briefly summarise the state-of-the-art) although not covering all threats on Machine Learning.