CL-Guard: Defending DNNs Against Backdoors via Fine-Grained Neuron Analysis and Collaborative Dual-Network Learning

摘要

Backdoor attacks on deep neural networks (DNNs) have garnered significant attention, particularly in edge computing applications. Given the complexity and opacity of DNNs, defending against backdoor attacks remains a formidable challenge. To address this, we propose CL-Guard, a dual-network-based defense framework designed to effectively eliminate potential backdoors in models. First, it leverages an inter-layer backpropagation algorithm to quantify each neuron's contribution to model prediction. Next, it constructs a critical neuron set through a recursive hierarchical partitioning method and an adaptive search strategy, identifying neurons critical to model prediction while minimizing the inclusion of backdoor-related neurons. Then, we perform sparse training on the non-critical neuron set, effectively strengthening the weights of critical neurons while disrupting the association between trigger features and backdoor-related neurons. Finally, we design a dual-network architecture that incorporates a fine-grained gradient backpropagation mechanism and dynamic collaborative learning, enabling the model to retain its original accuracy while preventing backdoor reactivation. The experimental results indicate that CL-Guard achieves an average Security Effectiveness Index (SEI) of approximately 95.42%, representing a 21.23% improvement over the state-of-the-art FT-SAM method.