Learning-based systems have been proved to be vulnerable against model inversion attacks (MIAs), where attackers steal private information of training data by querying the target model using synthetic samples. To alleviate the urgent threat introduced by MIAs, existing advancements are proposed to increase the attack overhead by limiting the information available. Although these methods successfully reduced the attack success rate (ASR) for a one-time inversion attempt, they usually compromise the usability of the protected model. More importantly, existing MIA defense methods fail to capture attack attempts, which can lead to persistent threats to data privacy. To bridge this gap, we propose Neural Honeypoint, an active defense framework against MIAs. The key insight is that MIA attackers will make a series of forward steps in the feature space while benign users will not. Motivated by the observation, defenders can deploy active defense devices (honeypoints) on critical paths to capture attack behaviors. Specifically, Neural Honeypoint first models the attackers’ capabilities from the frequency domain and designs specialized honeypoints for protected classes in the training dataset. Subsequently, it deploys these honeypoints into the protected model via backdoor-like model fine-tuning. Then, defenders can distinguish model inversion examples by comparing the similarity of input features with deployed honeypoints. Experiments show that Neural Honeypoint reduces the ASRs of advanced MIAs to 0%~2%. Furthermore, it can effectively capture inversion queries, which helps defenders to detect and block attacks in time.

Model Inversion Attacks (MIAs) pose a certain threat to the data privacy of learning-based systems, as they enable adversaries to reconstruct identifiable features of the training distribution with only query access to the victim model. In the context of deep learning, the primary challenges associated with MIAs are suboptimal attack success rates and the corresponding high computational costs. Prior efforts assumed that the expansive search space caused these limitations, employing generative models to constrain the dimensions of the search space. Despite the initial success of these generative-based solutions, recent experiments have cast doubt on this fundamental assumption, leaving two open questions about the influential factors determining MIA performance and how to manipulate these factors to improve MIAs. To answer these questions, we reframe MIAs from the perspective of information flow. This new formulation allows us to establish a lower bound for the error probability of MIAs, determined by two critical factors: (1) the size of the search space and (2) the mutual information between input and output random variables. Through a detailed analysis of generative-based MIAs within this theoretical framework, we uncover a trade-off between the size of the search space and the generation capability of generative models. Based on the theoretical conclusions, we introduce the Query-Efficient Model Inversion Approach (QE-MIA). By strategically selecting an appropriate search space and introducing additional mutual information, QE-MIA achieves a reduction of 60%~70% in query overhead while concurrently enhancing the attack success rate by 5%~25%.

Recent work has shown that deep neural networks are vulnerable to backdoor attacks. In comparison with the success of backdoor-attack methods, existing backdoor-defense methods face a lack of theoretical foundations and interpretable solutions. Most defense methods are based on experience with the characteristics of previous attacks, but fail to defend against new attacks. In this paper, we propose IBD, an interpretable backdoor-detection method via multivariate interactions. Using information theory techniques, IBD reveals how the backdoor works from the perspective of multivariate interactions of features. Based on the interpretable theorem, IBD enables defenders to detect backdoor models and poisoned examples without introducing additional information about the specific attack method. Experiments on widely used datasets and models show that IBD achieves a $78\%$ increase in average in detection accuracy and an order-of-magnitude reduction in time cost compared with existing backdoor-detection methods.

Despite the success of learning-based systems, recent studies have highlighted video adversarial examples as a ubiquitous threat to state-of-the-art video classification systems. Video adversarial attacks add subtle noise to the original example, resulting in a false classification result. Thorough studies on how to generate video adversarial examples are essential to prevent potential attacks. Despite much research on this, existing research works on the robustness of video adversarial examples are still limited. To generate highly robust video adversarial examples, we propose a video-augmentation-based adversarial attack (v3a), focusing on the video transformations to reinforce the attack. Further, we investigate different transformations as parts of the loss function to make the video adversarial examples more robust. The experiment results show that our proposed method outperforms other adversarial attacks in terms of robustness. We hope that our study encourages a deeper understanding of adversarial robustness in video classification systems with video augmentation.