# Robust Anti-Backdoor Instruction Tuning in LVLMs

arXiv: https://arxiv.org/abs/2506.05401 · HTML: https://mlsanity.com/paper/2506.05401/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 32 of its 41 references that are in the map.
- First posted on arXiv 4 Jun 2025.

## Placed in

- Multimodal LLMs / vision-language models / [MLLM jailbreaks and adversarial robustness](https://mlsanity.com/mllm/mllm-jailbreaks-and-adversarial-robustness/)
- 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 32)

- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [VQA: Visual Question Answering](https://mlsanity.com/paper/1505.00468/) (2015)
- [CIDEr: Consensus-based Image Description Evaluation](https://mlsanity.com/paper/1411.5726/) (2014)
- [Flickr30k Entities: Collecting Region-to-Phrase Correspondences for Richer Image-to-Sentence Models](https://mlsanity.com/paper/1505.04870/) (2015)
- [Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning](https://mlsanity.com/paper/1712.05526/) (2017)
- [Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks](https://mlsanity.com/paper/1805.12185/) (2018)
- [Spectral Signatures in Backdoor Attacks](https://mlsanity.com/paper/1811.00636/) (2018)
- [STRIP: A Defence Against Trojan Attacks on Deep Neural Networks](https://mlsanity.com/paper/1902.06531/) (2019)
- [Invisible Backdoor Attack with Sample-Specific Triggers](https://mlsanity.com/paper/2012.03816/) (2020)
- [Reflection Backdoor: A Natural Backdoor Attack on Deep Neural Networks](https://mlsanity.com/paper/2007.02343/) (2020)
- [MIMIC-IT: Multi-Modal In-Context Instruction Tuning](https://mlsanity.com/paper/2306.05425/) (2023)
- [A Comprehensive Survey of Deep Learning for Image Captioning](https://mlsanity.com/paper/1810.04020/) (2018)
- [LLaVA-Plus: Learning to Use Tools for Creating Multimodal Agents](https://mlsanity.com/paper/2311.05437/) (2023)
- [Backdoor Attacks and Countermeasures on Deep Learning: A Comprehensive Review](https://mlsanity.com/paper/2007.10760/) (2020)
- [BadCLIP: Dual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learning](https://mlsanity.com/paper/2311.12075/) (2023)
- [VL-Trojan: Multimodal Instruction Backdoor Attacks against Autoregressive Visual Language Models](https://mlsanity.com/paper/2402.13851/) (2024)
- [BadCLIP: Trigger-Aware Prompt Learning for Backdoor Attacks on CLIP](https://mlsanity.com/paper/2311.16194/) (2023)
- [Pre-trained Trojan Attacks for Visual Recognition](https://mlsanity.com/paper/2312.15172/) (2023)
- [Dual-Key Multimodal Backdoors for Visual Question Answering](https://mlsanity.com/paper/2112.07668/) (2021)
- [Poisoned Forgery Face: Towards Backdoor Attacks on Face Forgery Detection](https://mlsanity.com/paper/2402.11473/) (2024)
- [Compromising Embodied Agents with Contextual Backdoor Attacks](https://mlsanity.com/paper/2408.02882/) (2024)
- [Breaking the False Sense of Security in Backdoor Defense through Re-Activation Attack](https://mlsanity.com/paper/2405.16134/) (2024)
- [Towards Robust Physical-world Backdoor Attacks on Lane Detection](https://mlsanity.com/paper/2405.05553/) (2024)
- [Backdoor Attack through Frequency Domain](https://mlsanity.com/paper/2111.10991/) (2021)
- [Universal Backdoor Attacks Detection via Adaptive Adversarial Probe](https://mlsanity.com/paper/2209.05244/) (2022)
- [Defending Multimodal Backdoored Models by Repulsive Visual Prompt Tuning](https://mlsanity.com/paper/2412.20392/) (2024)

## Built on by (most cited first, 11 of 11)

- [Inject Once Survive Later: Backdooring Vision-Language-Action Models to Persist Through Downstream Fine-tuning](https://mlsanity.com/paper/2602.00500/) (2026)
- [BadCLIP++: Stealthy and Persistent Backdoors in Multimodal Contrastive Learning](https://mlsanity.com/paper/2602.17168/) (2026)
- [Self-Purification Mitigates Backdoors in Multimodal Diffusion Language Models](https://mlsanity.com/paper/2602.22246/) (2026)
- [Targeted Bit-Flip Attacks on LLM-Based Agents](https://mlsanity.com/paper/2603.10042/) (2026)
- [Hidden Ads: Behavior Triggered Semantic Backdoors for Advertisement Injection in Vision Language Models](https://mlsanity.com/paper/2603.27522/) (2026)
- [A Patch-based Cross-view Regularized Framework for Backdoor Defense in Multimodal Large Language Models](https://mlsanity.com/paper/2604.04488/) (2026)
- [EntropyScan: Towards Model-level Backdoor Detection in LVLMs via Visual Attention Entropy](https://mlsanity.com/paper/2605.15711/) (2026)
- [TimeGuard: Channel-wise Pool Training for Backdoor Defense in Time Series Forecasting](https://mlsanity.com/paper/2605.22365/) (2026)
- [TrojanWorld: Backdooring World-Model Agents via Imagination Steering](https://mlsanity.com/paper/2609.07051/) (2026)
- [TraceGuard: Adaptive Multimodal Poison Filtering through Cross-Feature Rank Agreement](https://mlsanity.com/paper/2609.29099/) (2026)
- [Selective Channel Restoration for Backdoored Vision-Language Models](https://mlsanity.com/paper/2609.37759/) (2026)

## Abstract (arXiv)

Large visual language models (LVLMs) have demonstrated excellent instruction-following capabilities, yet remain vulnerable to stealthy backdoor attacks when finetuned using contaminated data. Existing backdoor defense techniques are usually developed for single-modal visual or language models under fully parameter-adjustable settings or rely on supervisory knowledge during training. However, in real-world scenarios, defenders cannot modify frozen visual encoders or core LLM parameters, nor possess prior knowledge of unknown trigger patterns or target responses. Motivated by the empirical finding that LVLMs readily overfit to fixed, unknown triggers, which can embed malicious associations during adapter-level tuning, we aim to design a defense that operates without access to core weights or attack priors. To this end, we introduce a lightweight, certified-agnostic defense framework, Robust Instruction Tuning, that finetunes only adapter modules and text embedding layers under instruction tuning. Our method integrates two complementary regularizations: (1) Input Diversity Regularization, which perturbs trigger components across training samples to disrupt consistent spurious cues; and (2) Anomalous Activation Regularization, which dynamically sparses adapter weights exhibiting abnormally sharp activations linked to backdoor patterns. These mechanisms jointly guide the model toward learning semantically grounded representations rather than memorizing superficial trigger-response mappings. Extensive experiments against seven attacks on Flickr30k and MSCOCO demonstrate that ours reduces their attack success rate to nearly zero, with an increase in training cost of less than 15%.