# Bypassing Prompt Injection Detectors through Evasive Injections

arXiv: https://arxiv.org/abs/2602.00750 · HTML: https://mlsanity.com/paper/2602.00750/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Prompt injection and agent security, one of 12 directions in LLM safety and security.
- Builds on 12 of its 14 references that are in the map.
- First posted on arXiv 31 Jan 2026.

## Placed in

- LLM safety and security / [Prompt injection and agent security](https://mlsanity.com/safety/prompt-injection-and-agent-security/)

## Builds on (most cited first, 12 of 12)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Explaining and Harnessing Adversarial Examples](https://mlsanity.com/paper/1412.6572/) (2014)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://mlsanity.com/paper/2302.12173/) (2023)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [Active Retrieval Augmented Generation](https://mlsanity.com/paper/2305.06983/) (2023)
- [Formalizing and Benchmarking Prompt Injection Attacks and Defenses](https://mlsanity.com/paper/2310.12815/) (2023)
- [Get my drift? Catching LLM Task Drift with Activation Deltas](https://mlsanity.com/paper/2406.00799/) (2024)
- [Jedi: Entropy-based Localization and Removal of Adversarial Patches](https://mlsanity.com/paper/2304.10029/) (2023)
- [On Jailbreaking Quantized Language Models Through Fault Injection Attacks](https://mlsanity.com/paper/2507.03236/) (2025)
- [Attention Eclipse: Manipulating Attention to Bypass LLM Safety-Alignment](https://mlsanity.com/paper/2502.15334/) (2025)

## Built on by (most cited first, 1 of 1)

- [Readable Before Actionable: Causal Tracing of Indirect Prompt Injection](https://mlsanity.com/paper/2610.05295/) (2026)

## Abstract (arXiv)

Large language models (LLMs) are increasingly used in interactive and retrieval-augmented systems, but they remain vulnerable to prompt injection attacks, where injected secondary prompts force the model to deviate from the user's instructions to execute a potentially malicious task defined by the adversary. Recent work shows that ML models trained on activation shifts from LLMs' hidden layers can detect such drift. In this paper, we demonstrate that these detectors are not robust to adaptive adversaries. We propose a multi-probe evasion attack that appends an adversarially optimised suffix to poisoned inputs, jointly optimising a universal suffix to simultaneously fool all layer-wise drift detectors while preserving the effectiveness of the underlying injection. Using a modified Greedy Coordinate Gradient (GCG) approach, we generate universal suffixes that make prompt injections consistently evasive across multiple probes simultaneously. On Phi-3 3.8B and Llama-3 8B, a single suffix achieves attack success rates of 93.91% and 99.63% in successfully evading all detectors simultaneously. These results show that activation-based task drift detectors are highly vulnerable to adaptive prompt injection attacks, motivating stronger defences against such threats. We also propose a defence based on adversarial suffix augmentation: we generate multiple suffixes, append one at random during forward passes, and train detectors on the resulting activations. This approach is found to be effective against evasive attacks.