新方法通过消除触发器影响的对事实评分来改进代码模型的后门攻击

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-26 04:00

研究人员开发了一种名为消除触发器影响的对事实评分（TNCS）的新方法，以有效地识别用于投毒预训练代码模型的样本。该技术通过比较模型在触发输入与消除触发器影响的对事实输入上的输出，来隔离触发器的影响。TNCS 在非常低的投毒比例下，在各种代码任务和触发器类型上都展示了更高的攻击成功率，同时不会降低干净任务的性能。 AI

影响这项研究突显了代码模型的潜在漏洞，需要改进针对数据投毒攻击的防御措施。

排序理由该集群包含一篇研究论文，详细介绍了代码模型后门攻击的新方法。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

CodeXGLUE

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: 13 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇研究论文，详细介绍了代码模型后门攻击的新方法。[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: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Yu-Yu Lee, Ming-Feng Tsai, Tomohiro Morikawa, Chun-I Fan · 2026-10-26 04:00

通过触发器中性反事实评分实现数据高效的代码模型后门攻击

arXiv:2610.32334v1 Announce Type: cross Abstract: Pre-trained code models are exposed to backdoor poisoning, in which crafted training samples make the model misbehave on a trigger. A capable attacker minimizes the number of such samples, since each adds cost and exposure and a l…

本文如何被排名

报道来源 [1]

通过触发器中性反事实评分实现数据高效的代码模型后门攻击

相关话题