LeituraPesquisaRadarFramework de investimento
Entrar / Cadastrar
← Voltar para a leitura
NotíciasTópico · 幻影转移投毒攻击Material1 min de leituraEditorial DisconfirmAI

Data Poisoning Survives Filtering

New research shows 'Phantom Transfer' attacks evade 11 data-level defenses, even when the poisoning method is known.

2026-10-09 12:00 GMT+8

The 'Phantom Transfer' data poisoning attack evades 11 tested data-level defenses, including paraphrasing every sample.

The attack uses modified subliminal learning to embed malicious behaviors in training data. Even if developers know exactly how the poison was placed into an otherwise benign dataset, current filtering mechanisms cannot remove it.

Experiments demonstrate effectiveness regardless of the source model, target model, or attack goal. Authors suggest future defenses must supplement data cleaning with white-box methods and post-training audits.

This is a self-reported result from a NeurIPS 2026 accepted paper; the abstract does not specify whether tests were simulated or on real hardware.

Fontes

  • arxiv.org↗
Tópico

幻影转移投毒攻击

【背景】大型语言模型(LLM)依赖海量外部数据进行微调,这引入了数据投毒风险。传统观点认为,只要过滤掉可疑样本即可保障安全。然而,最新研究提出的“幻影转移”(Phantom Transfer)攻击证明这一假设已失效。

2 etapas1 peçasAtualizado 2026-10-09 14:18

Como chegamos aqui

  1. 2026-02-03
Entrar / Cadastrar
LeituraPesquisaRadarFramework de investimento
Andrew Draganov等人提交论文v1,首次提出“幻影转移”概念,指出数据级防御不足。
arxiv.org ↗
  • 2026-10-07论文更新至v3版本,被NeurIPS 2026接收,扩展了实验并证实攻击可绕过11种数据级防御。arxiv.org ↗
  • O que ainda está indefinido

    • 白盒审计工具能否有效识别由“幻影转移”植入的细微风格偏差?
    • 在实际生产环境中,针对此类攻击的模型后处理成本有多高?
    • 是否存在其他未测试的数据级防御组合能部分缓解此攻击?
    Ler o tópico completo →
    Continuar pesquisando →