OASIS: Mitigating Harmful Fine-tuning Attacks on LLMs via Orthogonal and Adaptive Safety Alignment Strategy
Year: 2026
Type: conference-paper
Abstract: The "Fine-Tuning-as-a-Service" paradigm exposes large language models to catastrophic safety degradation from less harmful samples.Alignment-stage defenses address this by proactively injecting adversarial perturbations to bolster the model's inherent robustness against harmful drift.However, existing methods rely on perturbation directions that often conflict with harmful gradients, inadvertently facilitating the acquisition of malicious features rather than suppressing them.To address this issue, we propose Orthogonal and Adaptive Safety Alignment Strategy (OASIS) to mathematically decouple safety enforcement from harmful feature acquisition.By projecting perturbations orthogonal to harmful gradients and concentrating optimization on adaptively selected safetycritical layers, OASIS effectively resolves directional conflicts while maximizing parameter efficiency.Extensive experiments on four LLMs across three datasets (SST2, GSM8K, and AGNews) demonstrate that OASIS reduces the Harmful Score by approximately 60% compared to competitive baselines, while maintaining stable downstream task utility.Our code is publicly available at https://github. com/xiaoroyi/OASIS.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Awards: G60673156, G7885342808
OpenAlex ID: W7166802823