arXiv Machine Learning August 20, 2026 By Shiyu Miao, Yunlong Mao, Zirui Huang, Liang Yao, Tianshuo Zheng, Yanhui Gu, Fan Liu, Sheng Zhong

Gradient Mirage: Trainable yet Label-Unidentifiable Gradients in Large Language Model Split Learning

Read the original on arXiv Machine Learning →

Gradient Mirage is a defense for large language model split learning that disrupts the assumption that exposed gradients faithfully reflect the client’s full-label training objective, thereby thwarting gradient matching attacks. It introduces inconsistencies across objective, direction, and scale by using selective autoregressive supervision, scale blinding, and directional privatization with a von Mises-Fisher mechanism. The method preserves utility through dual-track backpropagation and bottom-gradient recovery, achieving a stronger privacy‑utility trade‑off in experiments.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

arXiv AI

Jul 7

TIER: Trajectory-Invariant Explanation Regularization for Membership Privacy

arXiv:2607. 02903v1 Announce Type: cross Abstract: Explainability is central to building trustworthy AI, yet explanation interfaces can inadvertently provide adversaries with an expanded privacy-related attack surfaces.

By Varun Sharma, Kar Wai Fok, Vrizlynn L. L. Thing

safety

arXiv Machine Learning

3d ago

Approaching the Harm of Gradient Attacks While Only Flipping Labels

The paper investigates the impact of label‑flipping attacks on distributed machine learning, where an adversary can only flip a limited number of training labels. It formalizes the attack as a per‑round constrained optimization problem, derives a greedy label‑selection rule for logistic regression, and shows that this rule is provably optimal under mean aggregation. Experiments demonstrate that optimized label flipping can significantly degrade model accuracy, outperforming random flips, and that the attack transfers to other robust aggregators such as coordinate‑wise median and trimmed mean.

By Abdessamad El-Kabid, El-Mahdi El-Mhamdi

safety

arXiv Machine Learning

Jun 17

Loss Landscape Poisoning: Targeted Extraction of Unseen Training Data from LLMs

arXiv:2606. 17110v1 Announce Type: cross Abstract: Large Language Models are increasingly trained on proprietary or sensitive data, from private healthcare and financial records to user conversations containing secrets.

By Md Abdullah Al Mamun, Ngoc Phu Doan, Pedram Zaree, Ihsen Alouani, Nael Abu-Ghazaleh

llms multimodal safety

arXiv AI

Jun 24

Tuning without Peeking: Provable Generalization Bounds and Robust LLM Post-Training

arXiv:2507. 01752v4 Announce Type: replace-cross Abstract: Gradient-based optimization is the workhorse of deep learning, offering efficient and scalable training via backpropagation.

By Ismail Labiad, Mathurin Videau, Matthieu Kowalski, Marc Schoenauer, Alessandro Leite, Julia Kempe, Olivier Teytaud

llms benchmarks safety

arXiv Machine Learning

Aug 4

DeepDefense: Robust Learning via Layer-Wise Gradient-Feature Alignment

arXiv:2511. 13749v2 Announce Type: replace Abstract: Deep neural networks are known to be vulnerable to adversarial perturbations, which are small, carefully crafted inputs that lead to incorrect predictions.

By Ci Lin, Tet Yeap, Iluju Kiringa

safety

arXiv AI

Jun 11

Reinforcement Learning Disrupts Gradient-Based Adversarial Optimization

arXiv:2606. 12251v1 Announce Type: cross Abstract: Gradient-based adversarial attacks remain a dominant threat to deep neural networks (DNNs), as they exploit gradient information to efficiently optimize adversarial perturbations.

By Xinhai Zou, Chang Zhao, Alireza Aghabagherloo, Dave Singel\'ee, Robin Degraeve, Bart Preneel

reinforcement-learning safety

One email a morning, machine-written

Related stories

TIER: Trajectory-Invariant Explanation Regularization for Membership Privacy

Approaching the Harm of Gradient Attacks While Only Flipping Labels

Loss Landscape Poisoning: Targeted Extraction of Unseen Training Data from LLMs

Tuning without Peeking: Provable Generalization Bounds and Robust LLM Post-Training

DeepDefense: Robust Learning via Layer-Wise Gradient-Feature Alignment

Reinforcement Learning Disrupts Gradient-Based Adversarial Optimization