mlsanityData to 7 Oct 2026

InfoDecom: Decomposing Information for Defending Against Privacy Leakage in Split Inference

arXiv 2511.13365 ↗First posted 17 Nov 2025cs.CR cs.AI cs.DCcited by 0

Not on the map. Why?

Abstract (arXiv)

Split inference (SI) enables users to access deep learning (DL) services without directly transmitting raw data. However, recent studies reveal that data reconstruction attacks (DRAs) can recover the original inputs from the smashed data sent from the client to the server, leading to significant privacy leakage. While various defenses have been proposed, they often result in substantial utility degradation, particularly when the client-side model is shallow. We identify a key cause of this trade-off: existing defenses apply excessive perturbation to redundant information in the smashed data. To address this issue in computer vision tasks, we propose InfoDecom, a defense framework that first decomposes and removes redundant information and then injects noise calibrated to provide theoretically guaranteed privacy. Experiments demonstrate that InfoDecom achieves a superior utility-privacy trade-off compared to existing baselines.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 21 of its 40 references that are in the map. First posted on arXiv 17 Nov 2025.

Builds on

21 of 40 references.

References · 21

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  3. 2020Inverting Gradients -- How easy is it to break privacy in federated learning?cited by 707
  4. 2020CLUB: A Contrastive Log-ratio Upper Bound of Mutual Informationcited by 233
  5. 2016A guide to convolution arithmetic for deep learningcited by 182
  6. 2020Unleashing the Tiger: Inference Attacks on Split Learningcited by 104
  7. 2020Provable Defense against Privacy Leakage in Federated Learning from Representation Perspectivecited by 94
  8. 2022Semantic Communications for Future Internet: Fundamentals, Applications, and Challengescited by 94
  9. 2021CAFE: Catastrophic Data Leakage in Vertical Federated Learningcited by 76
  10. 2020NoPeek: Information leakage reduction to share activations in distributed deep learningcited by 66
  11. 2019Mitigating Information Leakage in Image Representations: A Maximum Entropy Approachcited by 48
  12. 2018Deep Private-Feature Extractioncited by 30
9 more
  1. 2022DuetFace: Collaborative Privacy-Preserving Face Recognition via Channel Splitting in the Frequency Domaincited by 18
  2. 2024Unveiling the Vulnerability of Private Fine-Tuning in Split-Based Frameworks for Large Language Models: A Bidirectionally Enhanced Attackcited by 17
  3. 2023Privacy Assessment on Reconstructed Images: Are Existing Evaluation Metrics Faithful to Human Perception?cited by 13
  4. 2023Task-Agnostic Privacy-Preserving Representation Learning for Federated Learning Against Attribute Inference Attackscited by 13
  5. 2024Inf2Guard: An Information-Theoretic Framework for Learning Privacy-Preserving Representations against Inference Attackscited by 13
  6. 2023Bounding the Invertibility of Privacy-preserving Instance Encoding using Fisher Informationcited by 11
  7. 2024Improving Robustness to Model Inversion Attacks via Sparse Coding Architecturescited by 8
  8. 2024Learning Robust and Privacy-Preserving Representations via Information Theorycited by 3
  9. 2025Quantifying Privacy Leakage in Split Inference via Fisher-Approximated Shannon Information Analysiscited by 2

Built on by

Not cited yet.