Advancements in Federated Learning Security: Our Research Presented at NDSS 2024
02/29/2024We are pleased to announce, that three of our research group's work on the security of Federated Learning systems has been accepted for presentation at the Network and Distributed System Security Symposium (NDSS) 2024, which is currently taking place in San Diego, California.
The papers of our team members, Torsten Krauß and Alexandra Dmitrienko, gained significant interest and attention from researchers and experts in the field at the Network and Distributed System Security Symposium (NDSS) 2024.
Among the presented papers there are two defenses against poisoning attacks in FL, namely CrowdGuard and FedFreq, as well as a tool to attack FL systems named AutoAdapt.
We are pleased by the opportunity to contribute to the advancement of the field and are delighted that our work has resonated within the cybersecurity research community.
Papers:
CrowdGuard: Federated Backdoor Detection in Federated Learning – A novel defense mechanism against backdoor attacks, ensuring the integrity of FL systems.
FreqFed: A Frequency Analysis-Based Approach for Mitigating Poisoning Attacks in Federated Learning – Introducing an innovative strategy to safeguard FL against poisoning attacks through frequency analysis.
Automatic Adversarial Adaption for Stealthy Poisoning Attacks in Federated Learning – A cutting-edge tool designed to test and improve the resilience of FL systems against adversarial threats.
CrowdGuard and FreqFed were developed in collaboration with the esteemed System Security Lab led by Professor Ahmad-Reza Sadeghi. This partnership has been instrumental in pushing the boundaries of what we can achieve in the realm of Federated Learning security.
We're proud to be at the forefront of developing solutions that ensure the security and reliability of FL systems worldwide.