Phillip Rieger (Technical University of Darmstadt), Thien Duc Nguyen (Technical University of Darmstadt), Markus Miettinen (Technical University of Darmstadt), Ahmad-Reza Sadeghi (Technical University of Darmstadt)

Federated Learning (FL) allows multiple clients to collaboratively train a Neural Network (NN) model on their private data without revealing the data. Recently, several targeted poisoning attacks against FL have been introduced. These attacks inject a backdoor into the resulting model that allows adversary-controlled inputs to be misclassified. Existing countermeasures against backdoor attacks are inefficient and often merely aim to exclude deviating models from the aggregation. However, this approach also removes benign models of clients with deviating data distributions, causing the aggregated model to perform poorly for such clients.

To address this problem, we propose emph{DeepSight}, a novel model filtering approach for mitigating backdoor attacks. It is based on three novel techniques that allow to characterize the distribution of data used to train model updates and seek to measure fine-grained differences in the internal structure and outputs of NNs. Using these techniques,
DeepSight can identify suspicious model updates. We also develop a scheme that can accurately cluster model updates. Combining the results of both components, DeepSight is able to identify and eliminate model clusters containing poisoned models with high attack impact. We also show that the backdoor contributions of possibly undetected poisoned models can be effectively mitigated with existing weight clipping-based defenses. We evaluate the performance and effectiveness of DeepSight and show that it can mitigate state-of-the-art backdoor attacks with a negligible impact on the model's performance on benign data.

View More Papers

ProvTalk: Towards Interpretable Multi-level Provenance Analysis in Networking Functions...

Azadeh Tabiban (CIISE, Concordia University, Montreal, QC, Canada), Heyang Zhao (CIISE, Concordia University, Montreal, QC, Canada), Yosr Jarraya (Ericsson Security Research, Ericsson Canada, Montreal, QC, Canada), Makan Pourzandi (Ericsson Security Research, Ericsson Canada, Montreal, QC, Canada), Mengyuan Zhang (Department of Computing, The Hong Kong Polytechnic University, China), Lingyu Wang (CIISE, Concordia University, Montreal, QC, Canada)

Read More

PoF: Proof-of-Following for Vehicle Platoons

Ziqi Xu (University of Arizona), Jingcheng Li (University of Arizona), Yanjun Pan (University of Arizona), Loukas Lazos (University of Arizona, Tucson), Ming Li (University of Arizona, Tucson), Nirnimesh Ghose (University of Nebraska–Lincoln)

Read More

EMS: History-Driven Mutation for Coverage-based Fuzzing

Chenyang Lyu (Zhejiang University), Shouling Ji (Zhejiang University), Xuhong Zhang (Zhejiang University & Zhejiang University NGICS Platform), Hong Liang (Zhejiang University), Binbin Zhao (Georgia Institute of Technology), Kangjie Lu (University of Minnesota), Raheem Beyah (Georgia Institute of Technology)

Read More

ROV-MI: Large-Scale, Accurate and Efficient Measurement of ROV Deployment

Wenqi Chen (Tsinghua University), Zhiliang Wang (Tsinghua University), Dongqi Han (Tsinghua University), Chenxin Duan (Tsinghua University), Xia Yin (Tsinghua University), Jiahai Yang (Tsinghua University), Xingang Shi (Tsinghua University)

Read More