Metraipolis › archive › dc_bf94d3b176b2
LoRAScan: Detecting Backdoor Prompts in Low-Rank Adapters for Large Language Models via Down-Projection Activation Spikes — arXiv:2608.06795
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.06795
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T09:47:09.667Z
- SHA-256 of the archived text
0070b0b31ed010128c4e433c626303fd01c56815c174d1d0ed499e2191029572- Stable URL of this page
https://metraipolis.com/documento/dc_bf94d3b176b2
How to cite
LoRAScan: Detecting Backdoor Prompts in Low-Rank Adapters for Large Language Models via Down-Projection Activation Spikes — arXiv:2608.06795 — arXiv (en), https://arxiv.org/abs/2608.06795 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bf94d3b176b2 (Metraipolis archive dc_bf94d3b176b2, sha256 0070b0b31ed010128c4e433c626303fd01c56815c174d1d0ed499e2191029572)
BibTeX
@misc{dcbf94d3b176b2,
title = {LoRAScan: Detecting Backdoor Prompts in Low-Rank Adapters for Large Language Models via Down-Projection Activation Spikes — arXiv:2608.06795},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.06795},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bf94d3b176b2; sha256 0070b0b31ed010128c4e433c626303fd01c56815c174d1d0ed499e2191029572},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.06795. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Low-rank adaptation (LoRA) enables efficient specialization and distribution of large language models through compact adapters. However, untrusted adapters introduce a supply-chain threat: a backdoored adapter can cause a model to generate harmful content, malicious code, political propaganda, or covert advertisements when an input contains a hidden trigger. Adapter-agnostic defenses merge the adapter with the base model, which dilutes backdoor signals and reduces detection performance. Existing adapter-aware methods do not address how to safely use a potentially backdoored adapter. Instead, they either train a defensive adapter to repair a backdoored base model, addressing the inverse problem rather than securing the adapter itself, or rely on a classifier that flags the entire adapter as suspicious and requires separate mitigation. These methods overlook the distinct latent-space signatures produced by trigger-bearing inputs in backdoored adapters. We introduce LoRAScan, the first adapter-aware defense that detects and rejects trigger-bearing inputs at inference time without modifying adapter parameters. Our key observation is that a small subset of LoRA insertion sites, approximately 5%, remains stable across clean inputs but exhibits highly concentrated spikes in LoRA down-projection activations when a trigger is present. LoRAScan identifies these low-variance insertion sites before model deployment and monitors them during inference. Across standard LLM backdoor benchmarks, LoRAScan rejects approximately 98.49 of malicious inputs with a small error rate on clean inputs, outperforming existing defenses across diverse evaluation settings. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-10 in cs.CL). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.06795 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-19T09:46:37.265Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-10?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).