Research
The research group is interested in the mathematics of cryptology, algorithmic number theory, evolutionary algorithms, computer vision and the study of convolutional neural networks.
Areas of research
Recent activity
Over the past few years, the team has been working on building adversarial images against CNNs. In particular, we have developed a black-box attack that is operational for several scenarios (targeted, untargeted, flat), and which is highly effective against all CNNs trained to classify images (on CIFAR-10, or on ImageNet for example) against which it was tested. We are currently working on similar attacks directly on high-resolution images, and on GANs