Our work, "A Synthetic Data-Driven Conformity Scoring Framework for Robust Federated Learning" will soon be presented at WACV 2026. It introduces a novel technique for robust federated learning, using synthetic data to defend against adversaries in a privacy-preserving way. Results show improved performance against gradient manipulation and backdoors. Our paper is available (https://openaccess.thecvf.com/content/WACV2026/papers/Alharbi_SD-CSFL_A_Synthetic_Data-Driven_Conformity_Scoring_Framework_for_Robust_Federated_WACV_2026_paper.pdf), including our code. Thanks E. Alharbi, A. Kerim, and Q. Ni! :) #ai #federatedlearning #wacv

My PhD student Mohammad Alsomali passed his viva last Tuesday! :) His thesis, "On-line Adaptation for Self-Adaptive Systems: Automatically Configuring Multi-arm Bandits Under Uncertainty", develops new techniques for on-line learning techniques to be applicable for automatically configuring computer systems (e.g., web-servers). Congratulations! Thanks Long Tran-Thanh, and Abdessalam Elhabbash for acting as examiners, and Phil Benachour for chairing the viva!

My student, Haoran Peng, successfully defended his thesis last Friday. His work, "Dynamic Early-Exiting Networks for Efficient Deep Learning", explores new techniques to improve the performance and applicability of efficient deep learning architectures. Congratulations for the hard work, and thanks Abdessalam Elhabbash, Valdir Grassi Jr, and Richard Jiang for conducting the viva! :)