VLALeaks: Membership Inference Attacks against Vision-Language-Action Models
This work addresses the underexplored privacy threat of membership inference in VLA models, which is critical for secure deployment in robotics.
The paper proposes VLALeaks, the first membership inference attack against Vision-Language-Action (VLA) models, exploiting attention discrepancies to determine if a sample was in the training set. Experiments show it achieves optimal attack AUC and TPR@1%FPR across multiple benchmarks, revealing privacy vulnerabilities in VLA models.
Vision-Language-Action (VLA) models enable end-to-end robot control and have garnered widespread attention. However, the memorization of training data inherent to VLA, coupled with the high cost of robotic data acquisition, raises serious concerns regarding data privacy leakage and intellectual property infringement. Membership inference attacks (MIAs) aim to determine whether a given sample belongs to the training set. While representing a significant privacy threat, this attack remains underexplored in the context of VLA models. To bridge this gap, we propose VLALeaks, which is based on attention discrepancies in VLA models. We reveal, for the first time, the privacy vulnerabilities of VLA models. Specifically, it comprises a two-stage process: (1) membership feature extraction, and (2) attack model construction. Experimental results across multiple VLA benchmarks demonstrate that VLALeaks readily reveals membership information and achieves optimal attack AUC and TPR@1\%FPR, highlighting the privacy vulnerabilities in current VLA model deployments. Our work is the first systematic study of MIAs on VLA models, aiming to provide insights for secure and trustworthy VLA models.