Variational Model Inversion Attacks

2022/01/26 by Kuan-Chieh Wang, Yan Fu, Wang, Kuan-Chieh +9 · 8 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · Medicine · #Adversarial Robustness in Machine Learning #Autopsy Techniques and Outcomes #Cryptography and Security (cs.CR) #FOS: Computer and information sciences #Forensic and Genetic Research #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2201.10787

openalex publication_date 2022/01/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Given the ubiquity of deep neural networks, it is important that these models do not reveal information about sensitive data that they have been trained on. In model inversion attacks, a malicious user attempts to recover the private dataset used to train a supervised neural network. A successful model inversion attack should generate realistic and diverse samples that accurately describe each of the classes in the private dataset. In this work, we provide a probabilistic interpretation of model inversion attacks, and formulate a variational objective that accounts for both diversity and accuracy. In order to optimize this variational objective, we choose a variational family defined in the code space of a deep generative model, trained on a public auxiliary dataset that shares some structural similarity with the target dataset. Empirically, our method substantially improves performance in terms of target attack accuracy, sample realism, and diversity on datasets of faces and chest X-ray images.

Cited by

• Rethinking Knowledge Distillation in Collaborative Machine Learning: Memory, Knowledge, and Their Interactions
• Model Inversion Attack Against Deep Hashing
• Model Inversion with Layer-Specific Modeling and Alignment for Data-Free Continual Learning
• Model Inversion Attacks Meet Cryptographic Fuzzy Extractors
• Rethinking Inference Placement for Deep Learning across Edge and Cloud Platforms: A Multi-Objective Optimization Perspective and Future Directions
• Generative Model Inversion Through the Lens of the Manifold Hypothesis
• Don't Trust the AI Ecosystem: Analyzing Privacy Leakage in Compromised Open-Source Components
• Accurate and Private Diagnosis of Rare Genetic Syndromes from Facial Images with Federated Deep Learning
• Do Vision-Language Models Leak What They Learn? Adaptive Token-Weighted Model Inversion Attacks

Related

• Plug & Play Attacks: Towards Robust and Flexible Model Inversion Attacks
• Re-thinking Model Inversion Attacks Against Deep Neural Networks
• The Secret Revealer: Generative Model-Inversion Attacks Against Deep Neural Networks
• Rethinking Model Inversion Attacks With Patch-Wise Reconstruction
• Model Inversion Attacks: A Survey of Approaches and Countermeasures