English(EN) Lilith: Backdoor Generalization under Training-Inference Trigger Shift

新的“Lilith”框架实现了机器学习模型的后门泛化

作者 PulseAugur 编辑部 · [1 个来源] · 2026-07-30 04:00

研究人员推出了一种名为Lilith的新框架，旨在解决机器学习模型的后门泛化问题。该框架专注于植入恶意行为，这些行为可以泛化到训练期间未见过的推理时触发器，这是现有后门研究在很大程度上忽略的一个漏洞。Lilith首先使用单个训练锚点创建一个紧凑的漏洞，然后构建一个保持表示几何的仅推理触发器族。实验表明，Lilith在实现高攻击成功率的同时，模型效用下降很小，触发器泛化差距也很小。 AI

影响突显了一类新的后门攻击，可能影响依赖于多样化数据源的机器学习服务的安全性。

排序理由详细介绍机器学习后门泛化新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍机器学习后门泛化新方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 107 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Zhou Feng, Jiahao Chen, Chunyi Zhou, Yuan Su, Tianyu Du, Yuwen Pu, Jianhai Chen, Jinbao Li, Shouling Ji · 2026-07-30 04:00

Lilith：训练-推理触发器偏移下的后门泛化

arXiv:2607.26099v1 Announce Type: cross Abstract: Machine-learning services increasingly rely on public data, third-party providers, and outsourced training, creating opportunities for data-poisoning attacks that implant persistent malicious behavior while preserving benign utili…

本文如何被排名

报道来源 [1]

Lilith：训练-推理触发器偏移下的后门泛化

相关实体

相关话题