External Projects

Internal Projects

  • Distribute ML: We are exploring the security and privacy dimensions of distributed machine learning. Specifically, we are developing novel attacks and defenses in prominent areas like Federated Learning and Split Learning. We aim to contribute valuable insights that enhance the robustness and resilience of distributed machine learning systems. Our focus encompasses the identification and mitigation of potential vulnerabilities, ensuring the confidentiality and integrity of data and models in these collaborative learning frameworks.
  • Hardware Fuzzing: Fuzzing's success in finding software vulnerabilities has led to its consideration for hardware security testing. Adapting software fuzzers for hardware faces challenges due to architectural differences. Our research focuses on creating an advanced hardware fuzzer for processors and SoCs. It employs information flow tracking and machine learning techniques to generate impactful test cases, offer detailed feedback, and identify subtle vulnerabilities such as information leaks.
  • ML and Physics Principles: Doppler Effects, Conservation of Energy, Conservation of Momentum or the Laws of Thermodynamics can be used to improve the results of some Artificial Intelligence neural networks? To answer this question, we are investigating the impact of this principles for XAI, text-detection or attacks defences.
  • Watermarking: NN training is a very expensive task, that consts companies billions of dollars to perform. To protect the intellectual property, it is possible to inject a watermarking inside the weights of the model (white-box) or on the output (black-box). We investigate how to inject and remove a watermark inside the model without losing accuracy.
  • GNN:We are currently investigating the use of Graph Neural Networks for Anomaly Detection and Hardware Fuzzing to improve the performances and speed up the process of those two tasks.
  • IoT: We take care of authentication, communication and more, everything with a special eye on security. We propose hands on projects, with PCB and IoT devices but we also analyse the security of IoT devices with Artificial Intelligence.
  • DeepFake Attacks and Protection: DeepFake and DeepFake Detector become more prevalent by the day. The creation of fake news is dangerous by virtue of making people confuse what’s true and what’s not. To combat this, we are currently investigating DeepFake to create attacks and countermeasures against it.