Towards a Robust, Effective, and Resource-Efficient Machine Learning Technique for IoT Security Monitoring

Join us for a seminar by Dr. Idris Zakariyya, a renowned cybersecurity professional, as he delves into the challenges and solutions for employing Deep Neural Networks in IoT security monitoring.

The prominence of Deep Neural Networks (DNNs) in IoT security monitoring has seen significant growth, but optimal detection performance remains an intricate puzzle due to computational constraints and adversarial susceptibilities. In this illuminating seminar, Dr. Idris Zakariyya presents a groundbreaking optimization method that merges regularization with simulated micro-batching. This innovative technique promises robust, efficient, and resource-conservative training of DNNs in IoT security contexts.

Key highlights include:

• A detailed overview of the challenges in implementing DNNs for IoT security.
• Introduction to the novel optimization method that amalgamates regularization and simulated micro-batching.
• Empirical results showcasing superior attack detection, resilience against adversarial disturbances, and notable reductions in memory and time consumption compared to conventional methods.
• A special focus on the model's performance in Federated Learning settings.

Speaker

Dr. Idris Zakariyya

Hosted by

Cybersecurity & Privacy, Department of Computing Science, University of Aberdeen

Venue

Hybrid (MacRobert MR815 & MS Teams)