Adversarial Attacks in Deep Neural Networks - Understanding Vulnerabilities and Countermeasures

Explore the intricacies of adversarial attacks in Machine Learning through this 21-minute conference talk delivered at Data Science Conference Europe 2023 in Belgrade. Delve into techniques used to deceive models through malicious inputs, particularly focusing on methods aimed at disrupting Deep Neural Networks (DNNs). Learn about the current challenges in defending against these attacks, as Aleksandar Tomcic examines the vulnerabilities of DNNs and discusses ongoing research efforts to develop effective countermeasures, highlighting that a definitive solution remains elusive.