# Defense Against LLM Backdoors using Critical Neuron Isolation Pruning

arXiv: https://arxiv.org/abs/2607.19894 · HTML: https://mlsanity.com/paper/2607.19894/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Backdoor and data poisoning attacks, one of 12 directions in LLM safety and security.
- Builds on 28 of its 45 references that are in the map.
- First posted on arXiv 22 Jul 2026.

## Placed in

- LLM safety and security / [Backdoor and data poisoning attacks](https://mlsanity.com/safety/backdoor-and-data-poisoning-attacks/)

## Builds on (most cited first, 28 of 28)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training](https://mlsanity.com/paper/2401.05566/) (2024)
- [Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions](https://mlsanity.com/paper/2309.07875/) (2023)
- [Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks](https://mlsanity.com/paper/2101.05930/) (2021)
- [ONION: A Simple and Effective Defense Against Textual Backdoor Attacks](https://mlsanity.com/paper/2011.10369/) (2020)
- [Adversarial Neuron Pruning Purifies Backdoored Deep Models](https://mlsanity.com/paper/2110.14430/) (2021)
- [Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection](https://mlsanity.com/paper/2307.16888/) (2023)
- [BadEdit: Backdooring large language models by model editing](https://mlsanity.com/paper/2403.13355/) (2024)
- [What Makes and Breaks Safety Fine-tuning? A Mechanistic Study](https://mlsanity.com/paper/2407.10264/) (2024)
- [CROW: Eliminating Backdoors from Large Language Models via Internal Consistency Regularization](https://mlsanity.com/paper/2411.12768/) (2024)
- [Learning Best Combination for Efficient N:M Sparsity](https://mlsanity.com/paper/2206.06662/) (2022)
- [Competition Report: Finding Universal Jailbreak Backdoors in Aligned LLMs](https://mlsanity.com/paper/2404.14461/) (2024)
- [QuSecNets: Quantization-based Defense Mechanism for Securing Deep Neural Network against Adversarial Attacks](https://mlsanity.com/paper/1811.01437/) (2018)
- [Do LLMs Give Psychometrically Plausible Responses in Educational Assessments?](https://mlsanity.com/paper/2506.09796/) (2025)
- [Simulate and Eliminate: Revoke Backdoors for Generative Large Language Models](https://mlsanity.com/paper/2405.07667/) (2024)
- [Obliviate: Neutralizing Task-agnostic Backdoors within the Parameter-efficient Fine-tuning Paradigm](https://mlsanity.com/paper/2409.14119/) (2024)
- [DPPA: Pruning Method for Large Language Model to Model Merging](https://mlsanity.com/paper/2403.02799/) (2024)
- [Opportunities and Challenges of LLMs in Education: An NLP Perspective](https://mlsanity.com/paper/2507.22753/) (2025)
- [P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs](https://mlsanity.com/paper/2510.04503/) (2025)
- [Pruning Strategies for Backdoor Defense in LLMs](https://mlsanity.com/paper/2508.20032/) (2025)
- [Benchmarking Ethical and Safety Risks of Healthcare LLMs in China-Toward Systemic Governance under Healthy China 2030](https://mlsanity.com/paper/2505.07205/) (2025)
- [MedicalOS: An LLM Agent based Operating System for Digital Healthcare](https://mlsanity.com/paper/2509.11507/) (2025)

## Abstract (arXiv)

Large language models (LLMs) are vulnerable to backdoor attacks, where hidden triggers induce malicious outputs. Existing defenses generally fall into inference-time detection or training-time mitigation, but face two key limitations. First, they focus on fine-tuning-based backdoors (e.g., PEFT modules) and fail to address insidious model-editing attacks that bypass training pipelines. Second, they target simple classification settings and do not naturally extend to open-ended LLM generation and do not naturally extend to the open-ended generation characteristics of LLMs. Consequently, these methods focus on surface-level behavioral patterns while neglecting the deeper representational causes of malicious activations. This lack of mechanistic understanding forces defenses to depend on empirical heuristics, limiting their robustness, generality, and practical applicability in real-world LLM deployment. To bridge this gap, we introduce DeCNIP (Defense with Critical Neuron Isolation Pruning), which leverages representational analysis to identify and neutralize backdoors in a unified pipeline. Specifically, DeCNIP identifies trigger-like behaviors by optimizing a cross-entropy loss between harmful prompts with candidate tokens and benign inputs. This representational discovery exposes latent threats by uncovering mechanisms through which triggers hijack model weights. It then isolates Backdoor Critical Neurons (BCNs) and prunes them selectively to remove malicious influence while preserving model utility. Extensive evaluations on six open-source LLMs and two benchmark datasets demonstrate that DeCNIP achieves over 95% relative reduction in Attack Success Rate (ASR), outperforming seven state-of-the-art defenses with only 0.1% neuron intervention. Moreover, it maintains 97% of the model's performance on normal benchmarks, demonstrating its efficacy, robustness, and scalability.