Certifying LLM intrusion detectors requires noise-augmented fine-tuning; restricting smoothing to attacker-controllable traffic features then gives 55–100% certified accuracy on two of three datasets.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Traffic-Aware Randomized Smoothing for LLM-Based Network Intrusion Detection
Certifying LLM intrusion detectors requires noise-augmented fine-tuning; restricting smoothing to attacker-controllable traffic features then gives 55–100% certified accuracy on two of three datasets.