Pith. sign in

REVIEW 11 cited by

BadEdit: Backdooring large language models by model editing

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2403.13355 v1 pith:SK2OTKPI submitted 2024-03-20 cs.CR cs.AI

BadEdit: Backdooring large language models by model editing

classification cs.CR cs.AI
keywords badeditbackdoorattackeditinginjectionmodelperformanceframework
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Mainstream backdoor attack methods typically demand substantial tuning data for poisoning, limiting their practicality and potentially degrading the overall performance when applied to Large Language Models (LLMs). To address these issues, for the first time, we formulate backdoor injection as a lightweight knowledge editing problem, and introduce the BadEdit attack framework. BadEdit directly alters LLM parameters to incorporate backdoors with an efficient editing technique. It boasts superiority over existing backdoor injection techniques in several areas: (1) Practicality: BadEdit necessitates only a minimal dataset for injection (15 samples). (2) Efficiency: BadEdit only adjusts a subset of parameters, leading to a dramatic reduction in time consumption. (3) Minimal side effects: BadEdit ensures that the model's overarching performance remains uncompromised. (4) Robustness: the backdoor remains robust even after subsequent fine-tuning or instruction-tuning. Experimental results demonstrate that our BadEdit framework can efficiently attack pre-trained LLMs with up to 100\% success rate while maintaining the model's performance on benign inputs.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Triggering Stealthy Feature Map Backdoors via Physical Fault Injection in Embedded Neural Networks

    cs.CR 2026-07 conditional novelty 7.5

    Precise electromagnetic or voltage faults on ARM Cortex-M4 can trigger feature-map backdoors in quantized CNNs that remain dormant without the fault and evade input-space detectors.

  2. Agent Security Bench (ASB): Formalizing and Benchmarking Attacks and Defenses in LLM-based Agents

    cs.CR 2024-10 unverdicted novelty 7.0

    ASB is a new benchmark that tests 10 prompt injection attacks, memory poisoning, a novel Plan-of-Thought backdoor attack, and 11 defenses on LLM agents across 13 models, finding attack success rates up to 84.3% and li...

  3. FORGE: Research-Trajectory Hijacking Attacks on Deep Research Agents

    cs.AI 2026-07 conditional novelty 6.5

    FORGE poisons deep-research planning with coordinated fake reasoning documents, reaching 26.4% PRISM report contamination at five injections; Root Query Anchoring halves that severity.

  4. Decision-Level Hijacking: Injecting Cognitive Bias into Large Language Models via Bit-Flip Attacks

    cs.CR 2026-07 conditional novelty 6.0

    A handful of weight-bit flips (as few as 12) can bias LLM outputs toward a target entity or stance, with limited effect on non-target tasks and output distributions.

  5. (A)iSpy: Parasitic Trojans for Machine Learning Infrastructure

    cs.CR 2026-07 conditional novelty 6.0

    A runtime-extension Trojan turns a single poisoned sample into a 97%+ backdoor via replay/amplification and leaks training hyperparameters through watermarked weights or innocuous text codewords.

  6. When the Aggregator Cheats: Data-Free Backdoors in Federated LLM-based QA Systems

    cs.CR 2026-06 unverdicted novelty 6.0

    Two-stage gradient-inversion attack recovers 5-20% of client samples to inject stealthy ad backdoors into federated QA LLMs, reaching ~100% ASR with negligible clean-task drop.

  7. Shared Latent Structures Enable Unified Backdoor Detection and Mitigation in LLMs

    cs.AI 2026-06 unverdicted novelty 6.0

    Sparse autoencoders identify shared latent features across diverse backdoor attacks in LLMs that enable unified detection via classifiers, causal control via steering, and mitigation via ablation fine-tuning.

  8. Defusing the Trigger: Plug-and-Play Defense for Backdoored LLMs via Tail-Risk Intrinsic Geometric Smoothing

    cs.CR 2026-04 unverdicted novelty 6.0

    TIGS detects backdoor-induced attention collapse in LLMs and applies content-aware tail-risk screening plus intrinsic geometric smoothing to suppress attacks while preserving normal performance.

  9. Stealthy Backdoor Attacks against LLMs Based on Natural Style Triggers

    cs.CR 2026-04 unverdicted novelty 6.0

    BadStyle creates stealthy backdoors in LLMs by poisoning samples with imperceptible style triggers and using an auxiliary loss to stabilize payload injection, achieving high attack success rates across multiple models...

  10. BackFlush: Knowledge-Free Backdoor Detection and Elimination with Watermark Preservation in Large Language Models

    cs.CR 2026-04 unverdicted novelty 6.0

    BackFlush detects backdoors via susceptibility amplification and eliminates them with RoPE unlearning to reach 1% ASR and 99% clean accuracy while preserving watermarks.

  11. Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey

    cs.CR 2024-09 unverdicted novelty 2.0

    Survey of harmful fine-tuning attacks on LLMs, their variants, defense strategies, mechanical analysis, and evaluation methodologies.