Research on Machine Unlearning Wins Best Paper Award at European Interdisciplinary Cybersecurity Conference 2024

Members of the Archimedes AI and Data Science Research Hub at the Athena Research Center received the Best Paper Award at the European Interdisciplinary Cybersecurity Conference (EICC) 2024 for their paper “ SFTC: Machine Unlearning via Selective Fine-Tuning and Targeted Confusion ” published in the ACM Digital Library.

The research team included Stathis Karypidis (Archimedes PhD scholar) and Nikos Komodakis (Assistant Professor, University of Crete, and Lead Researcher at Archimedes) in collaboration with Vasilis Perifanis (PhD scholar from Democritus University of Thrace and Adjunct researcher at Athena RC) and Pavlos Ephraimidis (Professor at Democritus University of Thrace, Greece and collaborating faculty member at Athena RC).

This award is a significant recognition of their efforts and contributions towards improving data security and privacy through Machine Unlearning. Machine Unlearning is a burgeoning area in artificial intelligence and data science that aims to remove specific data points from trained models without degrading their performance and without requiring complete retraining. Their research focuses on innovative methods that improve the efficiency and scalability of the unlearning process, making it crucial for enhancing data privacy and security. Through use of a Selective Fine-tuning and Targeted Confusion (SFTC) algorithm they show that SFTC preserves a model’s original accuracy while effectively inducing forgetting on the requested data samples.

The research team also recently received recognition at the international scientific competition "NeuIPS 2023 - Machine Unlearning," sponsored by Google – learn more: https://bit.ly/3zcs0dY

This success highlights the strength of interdisciplinary collaboration between Archimedes and Greek academic institutions and our continuous research efforts in the area of cybersecurity to address the complex challenges posed by modern data-driven technologies. Archimedes is dedicated to advancing the field of Machine Unlearning, contributing to a safer and more secure digital future.