🤖 AI Summary
This work addresses the vulnerability of large language models to traffic manipulation attacks in network intrusion detection by proposing a classifier-agnostic certified defense. The key innovation lies in restricting randomized smoothing to the attacker-controllable feature subspace and jointly optimizing robustness through noise-augmented fine-tuning and subspace alignment. This approach is the first to align the smoothing distribution with the actual threat model, overcoming the critical limitation of standard randomized smoothing—its extremely low certified accuracy on clean-trained models. Evaluated on the CIC-IDS-2018 and HIKARI-2021 datasets, the method achieves certified accuracies of 55%–100% and certified radii exceeding the L∞-equivalent thresholds by 1.8–5×, yielding up to a 72-percentage-point improvement over isotropic baselines.
📝 Abstract
Large language model (LLM)-based intrusion detection systems (IDS) are increasingly studied for security monitoring, yet their robustness against feasible traffic manipulation remains largely empirical. We present Traffic-Aware Randomized Smoothing (TA-RS), a classifier-agnostic certified defense that injects Gaussian noise exclusively into the directly controllable (DC) subspace -- features a remote attacker can modify -- during both fine-tuning and certification, aligning the smoothing distribution with the attacker-controllable subspace. We identify a critical prerequisite: applying standard randomized smoothing to clean-trained LLM-IDS yields weak certified accuracy in three of four (model, dataset) pairs tested (14-33%, at or below random) and only 57% in the fourth (43 pp below the noise-augmented result); noise-augmented fine-tuning recovers to 68-100% on two of three benchmark datasets (at sigma=0.25). At the L_inf-equivalent threshold R_inf = epsilon*sqrt(|DC|) (epsilon=0.05), TA-RS achieves 55-100% certified accuracy on CIC-IDS-2018 and HIKARI-2021, with median certified radii (R approx 0.45-0.96) exceeding R_inf by 1.8-5x (across sigma=0.25-1.00). Against a fairly trained iso-trained RS baseline the residual advantage is dataset-dependent (4-19 pp on CIC-IDS-2018). The larger gap -- up to 72 pp against an isotropic RS baseline that shares the DC-noise-augmented training recipe -- primarily reflects the training-certification mismatch rather than DC alignment alone: isotropic test-time noise perturbs uncontrollable features the attacker cannot exploit, triggering abstention rates up to 68%. RT-IoT2022 probes the limits of the method: it fails under the default fine-tuning recipe but recovers to 76%/69% certified accuracy (LLaMA3-8B/Qwen3-8B) when noise augmentation is increased.