About Session:

Generally, machine learning models for classification tasks are sensitive to adding noise. It is vulnerable to the perturbations intentionally crafted to mislead the classification result. The generation of such perturbations is called adversarial attack. The perturbed content is an adversarial example. As adversarial examples are an obvious threat, numerous methods for adversarial attacks and defences have been proposed in the literature. This special session promotes a challenging topic for developing generation and defence methods for adversarial attacks.