paper-with-me

홈 › Papers

A Failure-Mode Benchmark for Polymorphic Sybil Poisoning in RAG

2026-07-04 · Donghyun Lee, Juntae Kim arxiv

We release a benchmark and failure-mode-aware evaluation framework for grounded QA under coordinated retrieval poisoning. The framework partitions reader outputs into four mutually exclusive categories (\emph{gold}, \emph{hijack}, \emph{abstention}, \emph{drift}), with instance-level paired clean-to-poison transition matrices and a Forced Exposure protocol isolating reader-side conflict resolution from retrieval variance. We introduce \emph{polymorphic sybil poisoning}, a coordinated attack class in which $S$ lexically diverse passages jointly support an attacker-chosen target while evading lexical near-duplicate filters that fully detect monomorphic baselines (capturing the residual 14.2\% with E5 cosine raises false-positive rate 9$\times$ on legitimate same-topic pairs). A monomorphic-polymorphic ablation under Forced Exposure isolates the diversity dimension and reveals a $+$18.8pp hijack amplification (95\% paired bootstrap CI $[+15.4, +22.4]$, $B{=}5{,}000$): monomorphic copies register only 4.0\% as hijack while polymorphic surface diversity recovers 22.8\% -- a 5.7$\times$ amplification of the ASR-visible attack channel. ASR alone treats every non-target output identically; under attack, abstention and drift together hold 47-66\% of output mass, unmonitored by ASR+ACC, and two readers at nearly identical ASR (within 0.2pp) differ by 16.5pp on abstention and 17.2pp on drift -- failure profiles invisible to ASR. We release the frozen benchmark (3{,}145 questions, 2{,}982 retained sybil groups; $S{=}6$ chosen to dominate top-10 retrieval slots, §\ref{sec:setup}), the official four-way evaluator, paired-transition utilities, and the Forced Exposure harness across five readers (7B-120B), two retrievers, and two cross-validation datasets (TriviaQA, 2Wiki), under CC~BY-SA~4.0 (data) and MIT (software); release information in §\ref{sec:release}.

📄 PDF Abstract BibTeX arXiv:2607.03739

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Towards Sybil Resilience in Decentralized Learning

2023-06-26 · Thomas Werthenbach, Johan Pouwelse

Federated learning is a privacy-enforcing machine learning technology but suffers from limited scalability. This limitation mostly originates from the internet connection and memory capacity of the central parameter serv…

Adversarial AttackFederated Learning

Mitigating Sybils in Federated Learning Poisoning

2018-08-14 · Clement Fung, Chris J. M. Yoon, Ivan Beschastnikh

Machine learning (ML) over distributed multi-party data is required for a variety of domains. Existing approaches, such as federated learning, collect the outputs computed by a group of devices at a central aggregator an…

DiversityFederated LearningModel Poisoning

Sybil-based Virtual Data Poisoning Attacks in Federated Learning

2025-05-15 · Changxun Zhu, Qilong Wu, Lingjuan Lyu, Shibei Xue

Federated learning is vulnerable to poisoning attacks by malicious adversaries. Existing methods often involve high costs to achieve effective attacks. To address this challenge, we propose a sybil-based virtual data poi…

Data PoisoningFederated Learning

SaFL: Sybil-aware Federated Learning with Application to Face Recognition

2023-11-07 · Mahdi Ghafourian, Julian Fierrez, Ruben Vera-Rodriguez, Ruben Tolosana 외

Federated Learning (FL) is a machine learning paradigm to conduct collaborative learning among clients on a joint model. The primary goal is to share clients' local training parameters with an integrating server while pr…

Face RecognitionFederated Learning

Adversarial Attacks Against Deep Reinforcement Learning Framework in Internet of Vehicles

2021-08-02 · Anum Talpur, Mohan Gurusamy

Machine learning (ML) has made incredible impacts and transformations in a wide range of vehicular applications. As the use of ML in Internet of Vehicles (IoV) continues to advance, adversarial threats and their impact h…

Data PoisoningDeep Reinforcement Learningreinforcement-learningReinforcement Learning (RL)