PulseAugur
LIVE 06:29:09
research · [2 sources] ·
0
research

New research explores federated learning vulnerabilities and defenses against backdoor attacks

Researchers have developed new methods to combat sophisticated backdoor attacks in federated learning. One approach, DeTrigger, uses gradient analysis to detect and remove malicious triggers with minimal impact on model accuracy, achieving detection speeds up to 251x faster than traditional methods. Concurrently, another study introduced a Distributed Multi-Target Backdoor Attack (DMBA) framework that enables adversaries to control multiple clients with distinct triggers, demonstrating attack success rates above 80% for all implanted backdoors. AI

Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →

IMPACT New research highlights vulnerabilities in federated learning and proposes advanced defense mechanisms against sophisticated attacks.

RANK_REASON Two arXiv papers present novel methods for defending against and executing backdoor attacks in federated learning.

Read on arXiv cs.CV →

COVERAGE [2]

  1. arXiv cs.LG TIER_1 · Kichang Lee, Yujin Shin, Jonghyuk Yun, Songkuk Kim, Jun Han, JeongGil Ko ·

    DeTrigger: A Gradient-Centric Approach to Backdoor Attack Mitigation in Federated Learning

    arXiv:2411.12220v3 Announce Type: replace Abstract: Federated Learning (FL) enables collaborative model training across distributed devices while preserving local data privacy, making it ideal for mobile and embedded systems. However, the decentralized nature of FL also opens vul…

  2. arXiv cs.CV TIER_1 · Tao Liu, Dapeng Man, Jiguang Lv, Chen Xu, Weiye Xi, Huanran Wang, Yuhang Zhang, Tianming Zhao, Wu Yang ·

    Act in Collusion: Distributed Multi-Target Backdoor Attacks in Federated Learning

    arXiv:2411.03926v3 Announce Type: replace Abstract: Federated learning (FL) is widely used in Internet-of-Things (IoT) systems, but its distributed training process also exposes it to backdoor attacks. Existing studies mainly consider single-target or centralized multi-target set…