LoREnc: Low-Rank Encryption for Securing Foundation Models and LoRA Adapters

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ahn, Beomjin, Kwon, Jungmin, Jung, Chanyong, Chung, Jaewook
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916008711684096
author Ahn, Beomjin
Kwon, Jungmin
Jung, Chanyong
Chung, Jaewook
author_facet Ahn, Beomjin
Kwon, Jungmin
Jung, Chanyong
Chung, Jaewook
contents Foundation models and low-rank adapters enable efficient on-device generative AI but raise risks such as intellectual property leakage and model recovery attacks. Existing defenses are often impractical because they require retraining or access to the original dataset. We propose LoREnc, a training-free framework that secures both FMs and adapters via spectral truncation and compensation. LoREnc suppresses dominant low-rank components of FM weights, compensates for the missing information in authorized adapters, and further applies orthogonal reparameterization to obscure structural fingerprints of the protected adapter. Unauthorized users produce structurally collapsed outputs, while authorized users recover exact performance. Experiments demonstrate that LoREnc provides strong protection against model recovery with under 1% computational overhead.
format Preprint
id arxiv_https___arxiv_org_abs_2605_13163
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LoREnc: Low-Rank Encryption for Securing Foundation Models and LoRA Adapters
Ahn, Beomjin
Kwon, Jungmin
Jung, Chanyong
Chung, Jaewook
Cryptography and Security
Computer Vision and Pattern Recognition
Machine Learning
Foundation models and low-rank adapters enable efficient on-device generative AI but raise risks such as intellectual property leakage and model recovery attacks. Existing defenses are often impractical because they require retraining or access to the original dataset. We propose LoREnc, a training-free framework that secures both FMs and adapters via spectral truncation and compensation. LoREnc suppresses dominant low-rank components of FM weights, compensates for the missing information in authorized adapters, and further applies orthogonal reparameterization to obscure structural fingerprints of the protected adapter. Unauthorized users produce structurally collapsed outputs, while authorized users recover exact performance. Experiments demonstrate that LoREnc provides strong protection against model recovery with under 1% computational overhead.
title LoREnc: Low-Rank Encryption for Securing Foundation Models and LoRA Adapters
topic Cryptography and Security
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2605.13163