# Probe before You Talk: Towards Black-box Defense against Backdoor Unalignment for Large Language Models

arXiv: https://arxiv.org/abs/2506.16447 · HTML: https://mlsanity.com/paper/2506.16447/

## Key facts

- Cited by 19 papers in the map (data through 14 Oct 2026).
- Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 52 of its 64 references that are in the map.
- First posted on arXiv 19 Jun 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Safety alignment and jailbreak attacks](https://mlsanity.com/alignment/safety-alignment-and-jailbreak-attacks/)
- LLM safety and security / [Backdoor and data poisoning attacks](https://mlsanity.com/safety/backdoor-and-data-poisoning-attacks/)

## Builds on (most cited first, 30 of 52)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [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)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training](https://mlsanity.com/paper/2401.05566/) (2024)
- [DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models](https://mlsanity.com/paper/2306.11698/) (2023)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks](https://mlsanity.com/paper/2310.03684/) (2023)
- [Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation](https://mlsanity.com/paper/2310.06987/) (2023)
- [Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations](https://mlsanity.com/paper/2310.06387/) (2023)
- [Backdoor Learning: A Survey](https://mlsanity.com/paper/2007.08745/) (2020)
- [Weight Poisoning Attacks on Pre-trained Models](https://mlsanity.com/paper/2004.06660/) (2020)
- [ONION: A Simple and Effective Defense Against Textual Backdoor Attacks](https://mlsanity.com/paper/2011.10369/) (2020)
- [Hidden Killer: Invisible Textual Backdoor Attacks with Syntactic Trigger](https://mlsanity.com/paper/2105.12400/) (2021)
- [Mind the Style of Text! Adversarial and Backdoor Attacks Based on Text Style Transfer](https://mlsanity.com/paper/2110.07139/) (2021)
- [Tamper-Resistant Safeguards for Open-Weight LLMs](https://mlsanity.com/paper/2408.00761/) (2024)
- [Universal Jailbreak Backdoors from Poisoned Human Feedback](https://mlsanity.com/paper/2311.14455/) (2023)
- [Safe LoRA: the Silver Lining of Reducing Safety Risks when Fine-tuning Large Language Models](https://mlsanity.com/paper/2405.16833/) (2024)
- [Vaccine: Perturbation-aware Alignment for Large Language Models against Harmful Fine-tuning Attack](https://mlsanity.com/paper/2402.01109/) (2024)
- [Improved Techniques for Optimization-Based Jailbreaking on Large Language Models](https://mlsanity.com/paper/2405.21018/) (2024)
- [RAP: Robustness-Aware Perturbations for Defending against Backdoor Attacks on NLP Models](https://mlsanity.com/paper/2110.07831/) (2021)
- [SCALE-UP: An Efficient Black-box Input-level Backdoor Detection via Analyzing Scaled Prediction Consistency](https://mlsanity.com/paper/2302.03251/) (2023)
- [Lisa: Lazy Safety Alignment for Large Language Models against Harmful Fine-tuning Attack](https://mlsanity.com/paper/2405.18641/) (2024)
- [Booster: Tackling Harmful Fine-tuning for Large Language Models via Attenuating Harmful Perturbation](https://mlsanity.com/paper/2409.01586/) (2024)
- [Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey](https://mlsanity.com/paper/2409.18169/) (2024)

## Built on by (most cited first, 19 of 19)

- [BackdoorAgent: A Unified Framework for Backdoor Attacks on LLM-based Agents](https://mlsanity.com/paper/2601.04566/) (2026)
- [The Trigger in the Haystack: Extracting and Reconstructing LLM Backdoor Triggers](https://mlsanity.com/paper/2602.03085/) (2026)
- [Gradient Surgery for Safe LLM Fine-Tuning](https://mlsanity.com/paper/2508.07172/) (2025)
- [Backdoor Attribution: Elucidating and Controlling Backdoor in Language Models](https://mlsanity.com/paper/2509.21761/) (2025)
- [Watch the Weights: Unsupervised monitoring and control of fine-tuned LLMs](https://mlsanity.com/paper/2508.00161/) (2025)
- [Forgetting to Forget: Attention Sink as A Gateway for Backdooring LLM Unlearning](https://mlsanity.com/paper/2510.17021/) (2025)
- [Taught Well Learned Ill: Towards Distillation-conditional Backdoor Attack](https://mlsanity.com/paper/2509.23871/) (2025)
- [Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink](https://mlsanity.com/paper/2602.05228/) (2026)
- [PromptCOS: Towards Content-only System Prompt Copyright Auditing for LLMs](https://mlsanity.com/paper/2509.03117/) (2025)
- [From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs](https://mlsanity.com/paper/2510.05169/) (2025)
- [SLIP: Soft Label Mechanism and Key-Extraction-Guided CoT-based Defense Against Instruction Backdoor in APIs](https://mlsanity.com/paper/2508.06153/) (2025)
- [Diagnosing Knowledge Conflict in Multimodal Long-Chain Reasoning](https://mlsanity.com/paper/2602.14518/) (2026)
- [Depth Charge: Jailbreak Large Language Models from Deep Safety Attention Heads](https://mlsanity.com/paper/2603.05772/) (2026)
- [STEP: Detecting Audio Backdoor Attacks via Stability-based Trigger Exposure Profiling](https://mlsanity.com/paper/2603.18103/) (2026)
- [Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions](https://mlsanity.com/paper/2605.25073/) (2026)
- [TimeGuard: Channel-wise Pool Training for Backdoor Defense in Time Series Forecasting](https://mlsanity.com/paper/2605.22365/) (2026)
- [MASCRDM: Multi-Agent System for Compliance Risk Detection and Mitigation in Training Process of Large Language Models](https://mlsanity.com/paper/2609.39107/) (2026)
- [Localize-and-Detect: Auditing Task-Level Poisoning in Instruction-Tuned Models](https://mlsanity.com/paper/2610.03960/) (2026)
- [Towards Merging-conditional Backdoor Unalignment against Large Language Models](https://mlsanity.com/paper/2610.14701/) (2026)

## Abstract (arXiv)

Backdoor unalignment attacks against Large Language Models (LLMs) enable the stealthy compromise of safety alignment using a hidden trigger while evading normal safety auditing. These attacks pose significant threats to the applications of LLMs in the real-world Large Language Model as a Service (LLMaaS) setting, where the deployed model is a fully black-box system that can only interact through text. Furthermore, the sample-dependent nature of the attack target exacerbates the threat. Instead of outputting a fixed label, the backdoored LLM follows the semantics of any malicious command with the hidden trigger, significantly expanding the target space. In this paper, we introduce BEAT, a black-box defense that detects triggered samples during inference to deactivate the backdoor. It is motivated by an intriguing observation (dubbed the probe concatenate effect), where concatenated triggered samples significantly reduce the refusal rate of the backdoored LLM towards a malicious probe, while non-triggered samples have little effect. Specifically, BEAT identifies whether an input is triggered by measuring the degree of distortion in the output distribution of the probe before and after concatenation with the input. Our method addresses the challenges of sample-dependent targets from an opposite perspective. It captures the impact of the trigger on the refusal signal (which is sample-independent) instead of sample-specific successful attack behaviors. It overcomes black-box access limitations by using multiple sampling to approximate the output distribution. Extensive experiments are conducted on various backdoor attacks and LLMs (including the closed-source GPT-3.5-turbo), verifying the effectiveness and efficiency of our defense. Besides, we also preliminarily verify that BEAT can effectively defend against popular jailbreak attacks, as they can be regarded as 'natural backdoors'.