# CLIP-Inspector: Model-Level Backdoor Detection for Prompt-Tuned CLIP via OOD Trigger Inversion

arXiv: https://arxiv.org/abs/2604.09101 · HTML: https://mlsanity.com/paper/2604.09101/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 26 of its 45 references that are in the map.
- First posted on arXiv 10 Apr 2026.

## 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, 26 of 26)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Describing Textures in the Wild](https://mlsanity.com/paper/1311.3618/) (2013)
- [EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification](https://mlsanity.com/paper/1709.00029/) (2017)
- [Conditional Prompt Learning for Vision-Language Models](https://mlsanity.com/paper/2203.05557/) (2022)
- [BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain](https://mlsanity.com/paper/1708.06733/) (2017)
- [Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning](https://mlsanity.com/paper/1712.05526/) (2017)
- [STRIP: A Defence Against Trojan Attacks on Deep Neural Networks](https://mlsanity.com/paper/1902.06531/) (2019)
- [Hidden Trigger Backdoor Attacks](https://mlsanity.com/paper/1910.00033/) (2019)
- [Invisible Backdoor Attack with Sample-Specific Triggers](https://mlsanity.com/paper/2012.03816/) (2020)
- [Input-Aware Dynamic Backdoor Attack](https://mlsanity.com/paper/2010.08138/) (2020)
- [Universal Litmus Patterns: Revealing Backdoor Attacks in CNNs](https://mlsanity.com/paper/1906.10842/) (2019)
- [TABOR: A Highly Accurate Approach to Inspecting and Restoring Trojan Backdoors in AI Systems](https://mlsanity.com/paper/1908.01763/) (2019)
- [BadCLIP: Dual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learning](https://mlsanity.com/paper/2311.12075/) (2023)
- [Backdoor Scanning for Deep Neural Networks through K-Arm Optimization](https://mlsanity.com/paper/2102.05123/) (2021)
- [SCALE-UP: An Efficient Black-box Input-level Backdoor Detection via Analyzing Scaled Prediction Consistency](https://mlsanity.com/paper/2302.03251/) (2023)
- [Practical Detection of Trojan Neural Networks: Data-Limited and Data-Free Cases](https://mlsanity.com/paper/2007.15802/) (2020)
- [BadCLIP: Trigger-Aware Prompt Learning for Backdoor Attacks on CLIP](https://mlsanity.com/paper/2311.16194/) (2023)
- [Detecting Backdoors in Pre-trained Encoders](https://mlsanity.com/paper/2303.15180/) (2023)
- [Detecting Backdoors During the Inference Stage Based on Corruption Robustness Consistency](https://mlsanity.com/paper/2303.18191/) (2023)
- [UNICORN: A Unified Backdoor Trigger Inversion Framework](https://mlsanity.com/paper/2304.02786/) (2023)
- [Rethinking the Reverse-engineering of Trojan Triggers](https://mlsanity.com/paper/2210.15127/) (2022)
- [Defending Against Patch-based Backdoor Attacks on Self-Supervised Learning](https://mlsanity.com/paper/2304.01482/) (2023)
- [Backdoor Attack with Sparse and Invisible Trigger](https://mlsanity.com/paper/2306.06209/) (2023)
- [Detecting Backdoor Samples in Contrastive Language Image Pretraining](https://mlsanity.com/paper/2502.01385/) (2025)
- [TIJO: Trigger Inversion with Joint Optimization for Defending Multimodal Backdoored Models](https://mlsanity.com/paper/2308.03906/) (2023)
- [DeDe: Detecting Backdoor Samples for SSL Encoders via Decoders](https://mlsanity.com/paper/2411.16154/) (2024)

## Abstract (arXiv)

Organisations with limited data and computational resources increasingly outsource model training to Machine Learning as a Service (MLaaS) providers, who adapt vision-language models (VLMs) such as CLIP to downstream tasks via prompt tuning rather than training from scratch. This semi-honest setting creates a security risk where a malicious provider can follow the prompt-tuning protocol yet implant a backdoor, forcing triggered inputs to be classified into an attacker-chosen class, even for out-of-distribution (OOD) data. Such backdoors leave encoders untouched, making them undetectable to existing methods that focus on encoder corruption. Other data-level methods that sanitize data before training or during inference, also fail to answer the critical question, "Is the delivered model backdoored or not?" To address this model-level verification problem, we introduce CLIP-Inspector (CI), a backdoor detection method designed for prompt-tuned CLIP models. Assuming white-box access to the delivered model and a pool of unlabeled OOD images, CI reconstructs possible triggers for each class to determine if the model exhibits backdoor behaviour or not. Additionally, we demonstrate that using CI's reconstructed trigger for fine-tuning on correctly labeled triggered inputs enables us to re-align the model and reduce backdoor effectiveness. Through extensive experiments across ten datasets and four backdoor attacks, we demonstrate that CI can reconstruct effective triggers in a single epoch using only 1,000 OOD images, achieving a 94% detection accuracy (47/50 models). Compared to adapted trigger-inversion baselines, CI yields a markedly higher AUROC score (0.973 vs 0.495/0.687), thus enabling the vetting and post-hoc repair of prompt-tuned CLIP models to ensure safe deployment.