30 August 2023 in IEEE Transactions on Information Forensics and Security
Deep Neural Networks (DNNs) are vulnerable to adversarial patch attacks, which raises security concerns for face recognition systems using DNNs. Previous attack methods focus on the perturbation texture and generate adversarial patches with fixed shapes at random or pre-designed locations, which causes poor adversarial transferability. This paper proposes a Spatial Mutable Adversarial Patch (SMAP) method to generate a dynamic mutable patch to be injected into the face. In the proposed SMAP, the texture, position and shape of the patch are optimized simultaneously and the patch generation pipeline is end-to-end differentiable. Specifically, a Patch Location Selection Scheme is designed to find the critical patch position with the most significant influence on the target identity by the step-based gradient search. By innovatively bridging the pre-defined mask and the dynamic update of the patch, the patch position and shape are changed based on the affine transformation and sampling mechanism in each iteration, which maintains the importance of the injected patch to the adversarial objective. To evaluate the vulnerability of face recognition models, we explore more threatening impersonation attacks under the black-box setting and design a strict evaluation metric that aligns with the real-world scenario. Extensive experiments show that the proposed SMAP improves attack performance across various face recognition models and datasets. Moreover, SMAP achieves better transferability on commercial face recognition systems than existing methods.
3 April 2026 in IEEE Transactions on Circuits and Systems for Video Technology
Deep Neural Networks (DNNs) are vulnerable to adversarial patch attacks, which raises security concerns for face recognition systems using DNNs. Previous adversarial patch generation methods typically optimize perturbations in regions that maximally influence critical facial features. However, these existing methods are mostly limited to fixed shapes such as rectangles or squares. This confines subsequent patch texture optimization within these quadrilaterals, resulting in suboptimal adaptation to the complex geometric shapes of critical facial features, which may limit the effectiveness and transferability of the adversarial attacks. To address this issue, this paper proposes a PSO-based Adversarial Patch (PAP) method to generate a dynamic patch to be injected into the face. In the proposed PAP, by employing Particle Swarm Optimization (PSO) with adversarial similarity as the objective, the algorithm searches within a base circle to determine the optimal shape and position of the pre-defined patch. This approach enables the patch to exhibit extrapolation of polygonal deformations, ensuring that the patch optimally balances location, texture, and geometry, which enhances the adversarial transferability of the patch. To evaluate the vulnerability of face recognition models, we explore impersonation attacks under the black-box setting. Extensive experiments show that the proposed PAP improves attack performance across various face recognition models and datasets. Moreover, PAP achieves better transferability on commercial face recognition systems than existing methods.