Search Schedule From Shared Calendar. ChikkuTakku is service to search schedule from shared calendar and help share calendar
Speaker: Sabre Didi Title: Co-evolutionary Adversarial Learning for Robust Malware DetectionAbstract: Adversarial attacks pose a growing challenge to machine learning systems used in malware detection, particularly when models rely on complex, high-dimensional feature representations such as those found in the EMBER dataset. In this talk, I present a co-evolutionary adversarial training framework designed to improve the robustness of such systems. The approach combines gradient-based adversarial attacks with evolutionary optimisation to create adaptive and increasingly challenging adversaries. A hybrid CNN–MLP model is used to capture both structural patterns and metadata features within malware samples. Adversarial examples are generated using Projected Gradient Descent (PGD); however, rather than relying on fixed parameters, an evolutionary controller dynamically adjusts the attack strategy based on gradient information. This results in an online co-evolutionary process in which the classifier and adversary continuously adapt to one another, mirroring the real-world arms race in malware detection. Experimental results show that this method improves robustness against adversarial attacks while maintaining strong performance on clean data. This work highlights how combining deep learning with evolutionary strategies can lead to more resilient cybersecurity systems and provides insight into the future of adaptive defence mechanisms in adversarial environments.Biography:Dr. Sabre Z. Didi completed his PhD (2018) in Computer Science and undertook postdoctoral research (2019–2022) in the evolutionary machine learning research group under the supervision of Professor Geoff Nitschke at the University of Cape Town (UCT), specialising in Evolutionary (Robotic) Controller Design. Between 2018 and 2025, he worked in the software development industry across multiple organisations, holding roles of increasing responsibility, including positions in AI engineering and software development. Notably, he worked as an AI and Software Engineer at Sedna, a UK-based company specialising in the maritime sector. He is currently a Lecturer at the Cape Peninsula University of Technology in the Department of Information Technology.
📍 John Day LT1