While existing defenses often rely on regularization techniques to reduce information leakage, they remain vulnerable to recent attacks.

In model inversion attacks (MIAs), adversaries attempt to recover the private training data by exploiting access to a well-trained target model.

The wide adoption of deep neural networks (DNNs) in mission-critical applications has spurred the need for interpretable models that provide explanations of the model's decisions. Unfortunately, previous studies have demonstrated that model explanations facilitate information leakage, rendering DNN models vulnerable to model inversion attacks. These attacks enable the adversary to reconstruct original images based on model explanations, thus leaking privacy-sensitive features. To this end, we present Generative Noise Injector for Model Explanations (GNIME), a novel defense framework that perturbs model explanations to minimize the risk of model inversion attacks while preserving the interpretabilities of the generated explanations. Specifically, we formulate the defense training as a two-player minimax game between the inversion attack network on the one hand, which aims to invert model explanations, and the noise generator network on the other, which aims to inject perturbations to tamper with model inversion attacks. We demonstrate that GNIME significantly decreases the information leakage in model explanations, decreasing transferable classification accuracy in facial recognition models by up to 84.8% while preserving the original functionality of model explanations.

Model inversion (MI) attacks pose significant privacy risks by reconstructing private training data from trained neural networks. While prior studies have primarily examined unimodal deep networks, the vulnerability of vision-language models (VLMs) remains largely unexplored. In this work, we present the first systematic study of MI attacks on VLMs to understand their susceptibility to leaking private visual training data. Our work makes two main contributions. First, tailored to the token-generative nature of VLMs, we introduce a suite of token-based and sequence-based model inversion strategies, providing a comprehensive analysis of VLMs' vulnerability under different attack formulations. Second, based on the observation that tokens vary in their visual grounding, and hence their gradients differ in informativeness for image reconstruction, we propose Sequence-based Model Inversion with Adaptive Token Weighting (SMI-AW) as a novel MI for VLMs. SMI-AW dynamically reweights each token's loss gradient according to its visual grounding, enabling the optimization to focus on visually informative tokens and more effectively guide the reconstruction of private images. Through extensive experiments and human evaluations on a range of state-of-the-art VLMs across multiple datasets, we show that VLMs are susceptible to training data leakage. Human evaluation of the reconstructed images yields an attack accuracy of 61.21%, underscoring the severity of these privacy risks. Notably, we demonstrate that publicly released VLMs are vulnerable to such attacks. Our study highlights the urgent need for privacy safeguards as VLMs become increasingly deployed in sensitive domains such as healthcare and finance. Additional experiments are provided in Supp.

Deep hashing improves retrieval efficiency through compact binary codes, yet it introduces severe and often overlooked privacy risks. The ability to reconstruct original training data from hash codes could lead to serious threats such as biometric forgery and privacy breaches. However, model inversion attacks specifically targeting deep hashing models remain unexplored, leaving their security implications unexamined. This research gap stems from the inaccessibility of genuine training hash codes and the highly discrete Hamming space, which prevents existing methods from adapting to deep hashing. To address these challenges, we propose DHMI, the first diffusion-based model inversion framework designed for deep hashing. DHMI first clusters an auxiliary dataset to derive semantic hash centers as surrogate anchors. It then introduces a surrogate-guided denoising optimization method that leverages a novel attack metric (fusing classification consistency and hash proximity) to dynamically select candidate samples. A cluster of surrogate models guides the refinement of these candidates, ensuring the generation of high-fidelity and semantically consistent images. Experiments on multiple datasets demonstrate that DHMI successfully reconstructs high-resolution, high-quality images even under the most challenging black-box setting, where no training hash codes are available. Our method outperforms the existing state-of-the-art model inversion attacks in black-box scenarios, confirming both its practical efficacy and the critical privacy risks inherent in deep hashing systems.

Model Inversion (MI) attacks aim to reconstruct information from private training data by exploiting access to machine learning models T. To evaluate such attacks, the standard evaluation framework relies on an evaluation model E, trained under the same task design as T. This framework has become the de facto standard for assessing progress in MI research, used across nearly all recent MI studies without question. In this paper, we present the first in-depth study of this evaluation framework. In particular, we identify a critical issue of this standard framework: Type-I adversarial examples. These are reconstructions that do not capture the visual features of private training data, yet are still deemed successful by T and ultimately transferable to E. Such false positives undermine the reliability of the standard MI evaluation framework. To address this issue, we introduce a new MI evaluation framework that replaces the evaluation model E with advanced Multimodal Large Language Models (MLLMs). By leveraging their general-purpose visual understanding, our MLLM-based framework does not depend on training of shared task design as in T, thus reducing Type-I transferability and providing more faithful assessments of reconstruction success. Using our MLLM-based evaluation framework, we reevaluate 27 diverse MI attack setups and empirically reveal consistently high false positive rates under the standard evaluation framework. Importantly, we demonstrate that many state-of-the-art (SOTA) MI methods report inflated attack accuracy, indicating that actual privacy leakage is significantly lower than previously believed. By uncovering this critical issue and proposing a robust solution, our work enables a reassessment of progress in MI research and sets a new standard for reliable and robust evaluation. Code can be found in https://github.com/hosytuyen/MI-Eval-MLLM

Model inversion attacks pose an open challenge to privacy-sensitive applications that use machine learning (ML) models. For example, face authentication systems use modern ML models to compute embedding vectors from face images of the enrolled users and store them. If leaked, inversion attacks can accurately reconstruct user faces from the leaked vectors. There is no systematic characterization of properties needed in an ideal defense against model inversion, even for the canonical example application of a face authentication system susceptible to data breaches, despite a decade of best-effort solutions. In this paper, we formalize the desired properties of a provably strong defense against model inversion and connect it, for the first time, to the cryptographic concept of fuzzy extractors. We further show that existing fuzzy extractors are insecure for use in ML-based face authentication. We do so through a new model inversion attack called PIPE, which achieves a success rate of over 89% in most cases against prior schemes. We then propose L2FE-Hash, the first candidate fuzzy extractor which supports standard Euclidean distance comparators as needed in many ML-based applications, including face authentication. We formally characterize its computational security guarantees, even in the extreme threat model of full breach of stored secrets, and empirically show its usable accuracy in face authentication for practical face distributions. It offers attack-agnostic security without requiring any re-training of the ML model it protects. Empirically, it nullifies both prior state-of-the-art inversion attacks as well as our new PIPE attack.

While existing defenses often rely on regularization techniques to reduce information leakage, they remain vulnerable to recent attacks.