How Can We Achieve Efficient Privacy-Preserving Machine Learning? A Cross‑Layer Review

This systematic review, authored by Peking University and Ant Group researchers, examines efficient privacy‑preserving machine learning (PPML) across protocol, model, and system layers, identifies key optimization challenges, proposes a unified cross‑layer perspective, and outlines future research directions for large‑model privacy and performance.

Cross-Layer OptimizationEfficient MLMachine Learning Security

0 likes · 7 min read

How Can We Achieve Efficient Privacy-Preserving Machine Learning? A Cross‑Layer Review

Alimama Tech

Sep 20, 2023 · Artificial Intelligence

Exploring Model Dynamics for Accumulative Poisoning Detection

The paper, a joint effort by Alibaba Mama and HKBU TMLR, shows that monitoring model dynamics—specifically a newly defined memorization‑discrepancy metric—can reveal hidden accumulative poisoning attacks in online advertising streams, and introduces a discrepancy‑aware correction algorithm that consistently outperforms existing defenses across benchmark datasets.

Machine Learning SecurityOnline Learningdefense algorithms

0 likes · 13 min read

Exploring Model Dynamics for Accumulative Poisoning Detection

AntTech

Oct 31, 2022 · Artificial Intelligence

Automated Attacker A² for Enhancing Model Robustness in Adversarial Training

The paper presents A², an automated, parameterized attacker that dynamically adjusts perturbation methods and step sizes during adversarial training, demonstrating improved robustness across multiple benchmarks with modest computational overhead, and outlines future directions for further efficiency and effectiveness in secure AI systems.

Machine Learning SecurityNeurIPSadversarial training

0 likes · 9 min read

Automated Attacker A² for Enhancing Model Robustness in Adversarial Training