mlsanityData to 7 Oct 2026

Label Leakage and Protection in Two-party Split Learning

arXiv 2102.08504 ↗First posted 17 Feb 2021cs.LG cs.CRcited by 87

Not on the map. Why?

Abstract (arXiv)

Two-party split learning is a popular technique for learning a model across feature-partitioned data. In this work, we explore whether it is possible for one party to steal the private label information from the other party during split training, and whether there are methods that can protect against such attacks. Specifically, we first formulate a realistic threat model and propose a privacy loss metric to quantify label leakage in split learning. We then show that there exist two simple yet effective methods within the threat model that can allow one party to accurately recover private ground-truth labels owned by the other party. To combat these attacks, we propose several random perturbation techniques, including textttMarvell, an approach that strategically finds the structure of the noise perturbation by minimizing the amount of label leakage (measured through our quantification metric) of a worst-case adversary. We empirically demonstrate the effectiveness of our protection techniques against the identified attacks, and show that textttMarvell in particular has improved privacy-utility tradeoffs relative to baseline approaches.

Key facts. Cited by 87 papers in the map (data through 7 Oct 2026). Builds on 20 of its 44 references that are in the map. First posted on arXiv 17 Feb 2021.

Builds on

20 of 44 references.

References · 20

  1. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
  2. 2016Deep Learning with Differential Privacycited by 2,682
  3. 2019Deep Leakage from Gradientscited by 1,004
  4. 2016Wide & Deep Learning for Recommender Systemscited by 745
  5. 2017Differentially Private Federated Learning: A Client Level Perspectivecited by 451
  6. 2019Measuring Calibration in Deep Learningcited by 443
  7. 2018Split learning for health: Distributed deep learning without sharing raw patient datacited by 391
  8. 2020iDLG: Improved Deep Leakage from Gradientscited by 367
  9. 2018Distributed learning of deep neural network over multiple agentscited by 277
  10. 2019Differentially Private Learning with Adaptive Clippingcited by 218
  11. 2020Threats to Federated Learning: A Surveycited by 171
  12. 2018Amplification by Shuffling: From Local to Central Differential Privacy via Anonymitycited by 163
8 more
  1. 2018Protection Against Reconstruction and Its Applications in Private Federated Learningcited by 127
  2. 2018Distributed Differential Privacy via Shufflingcited by 115
  3. 2021Deep Learning with Label Differential Privacycited by 113
  4. 2018Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardwarecited by 101
  5. 2018A General Approach to Adding Differential Privacy to Iterative Training Procedurescited by 84
  6. 2019AdaCliP: Adaptive Clipping for Private SGDcited by 77
  7. 2019QUOTIENT: Two-Party Secure Neural Network Training and Predictioncited by 41
  8. 2018A Review of Homomorphic Encryption Libraries for Secure Computationcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 7; 2022: 20; 2023: 25; 2024: 18; 2025: 10; 2026: 7.

87 papers.

Sort:

By area:

  1. 2022Vertical Federated Learning: Concepts, Advances and Challengescited by 104
  2. 2022Survey on Federated Learning Threats: concepts, taxonomy on attacks and defences, experimental study and challengescited by 81
  3. 2021UnSplit: Data-Oblivious Model Inversion, Model Stealing, and Label Inference Attacks Against Split Learningcited by 52
  4. 2022BlindFL: Vertical Federated Machine Learning without Peeking into Your Datacited by 36
  5. 2022Trading Off Privacy, Utility and Efficiency in Federated Learningcited by 33
  6. 2023A Survey on Vertical Federated Learning: From a Layered Perspectivecited by 30
  7. 2021Privacy-preserving Federated Adversarial Domain Adaption over Feature Groups for Interpretabilitycited by 28
  8. 2022Label Leakage and Protection from Forward Embedding in Vertical Federated Learningcited by 26
  9. 2022Differentially Private Label Protection in Split Learningcited by 25
  10. 2021User-Level Label Leakage from Gradients in Federated Learningcited by 21
  11. 2021Vertical Federated Learning without Revealing Intersection Membershipcited by 19
  12. 2021Graph-Fraudster: Adversarial Attacks on Graph Neural Network Based Vertical Federated Learningcited by 17
38 more
  1. 2022A Hybrid Self-Supervised Learning Framework for Vertical Federated Learningcited by 17
  2. 2023Privacy-Aware Document Visual Question Answeringcited by 17
  3. 2021Batch Label Inference and Replacement Attacks in Black-Boxed Vertical Federated Learningcited by 16
  4. 2021ExPLoit: Extracting Private Labels in Split Learningcited by 15
  5. 2022Similarity-based Label Inference Attack against Training and Inference of Split Learningcited by 15
  6. 2022Towards Communication-efficient Vertical Federated Learning Training via Cache-enabled Local Updatescited by 15
  7. 2023On the Trustworthiness Landscape of State-of-the-art Generative Models: A Survey and Outlookcited by 15
  8. 2022Combined Federated and Split Learning in Edge Computing for Ubiquitous Intelligence in Internet of Things: State of the Art and Future Directionscited by 14
  9. 2022Regression with Label Differential Privacycited by 14
  10. 2024Vertical Federated Learning for Effectiveness, Security, Applicability: A Surveycited by 14
  11. 2022Feature Reconstruction Attacks and Countermeasures of DNN training in Vertical Federated Learningcited by 13
  12. 2022Residue-based Label Protection Mechanisms in Vertical Logistic Regressioncited by 10
  13. 2022A Framework for Evaluating Privacy-Utility Trade-off in Vertical Federated Learningcited by 10
  14. 2023FedPass: Privacy-Preserving Vertical Federated Deep Learning with Adaptive Obfuscationcited by 9
  15. 2023Client-specific Property Inference against Secure Aggregation in Federated Learningcited by 9
  16. 2023Passive Inference Attacks on Split Learning via Adversarial Regularizationcited by 9
  17. 2023Federated Learning of Models Pre-Trained on Different Features with Consensus Graphscited by 8
  18. 2023VFLAIR: A Research Library and Benchmark for Vertical Federated Learningcited by 8
  19. 2022Is Vertical Logistic Regression Privacy-Preserving? A Comprehensive Privacy Analysis and Beyondcited by 7
  20. 2022DPAUC: Differentially Private AUC Computation in Federated Learningcited by 7
  21. 2023Label Inference Attack against Split Learning under Regression Settingcited by 7
  22. 2023Robust and IP-Protecting Vertical Federated Learning against Unexpected Quitting of Partiescited by 7
  23. 2024A Survey of Privacy Threats and Defense in Vertical Federated Learning: From Model Life Cycle Perspectivecited by 7
  24. 2025SafeSplit: A Novel Defense Against Client-Side Backdoor Attacks in Split Learning (Full Version)cited by 7
  25. 2023Universal Adversarial Backdoor Attacks to Fool Vertical Federated Learning in Cloud-Edge Collaborationcited by 6
  26. 2023FedAds: A Benchmark for Privacy-Preserving CVR Estimation with Vertical Federated Learningcited by 6
  27. 2023Eliminating Label Leakage in Tree-Based Vertical Federated Learningcited by 6
  28. 2023Optimal Unbiased Randomizers for Regression with Label Differential Privacycited by 6
  29. 2024A Hybrid Quantum Neural Network for Split Learningcited by 5
  30. 2023FedLAP-DP: Federated Learning by Sharing Differentially Private Loss Approximationscited by 4
  31. 2023SplitOut: Out-of-the-Box Training-Hijacking Detection in Split Learning via Outlier Detectioncited by 4
  32. 2023PrivaScissors: Enhance the Privacy of Collaborative Inference through the Lens of Mutual Informationcited by 4
  33. 2023Analyzing the vulnerabilities in SplitFed Learning: Assessing the robustness against Data Poisoning Attackscited by 4
  34. 2025VFLAIR-LLM: A Comprehensive Framework and Benchmark for Split Learning of LLMscited by 4
  35. 2022Differentially Private AUC Computation in Vertical Federated Learningcited by 3
  36. 2022Measuring and Controlling Split Layer Privacy Leakage Using Fisher Informationcited by 3
  37. 2022Protecting Split Learning by Potential Energy Losscited by 3
  38. 2023Mutual Information Regularization for Vertical Federated Learningcited by 3

50 most cited of 87 All 87 (JSON)