Abstract
Proposed herein is a system that provides a training methodology for large language models (LLMs) that classifies fully encrypted network traffic into Quality of Service (QoS)-relevant types using only Layer 3 and Layer 4 statistical features, including packet sizes, timing, throughput, and directionality, from the first 100 packets of each flow. The system operates entirely on observable behavioral patterns and currently achieves 65-68% accuracy, with a path toward at least 85% accuracy through purpose-built network LLMs and expanded datasets. The system also provides explainable synthetic reasoning, low-rank adaptation (LoRA)-based incremental learning, and classification latency below 100 milliseconds for QoS enforcement in fully encrypted networks.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Subramanian, Rajasekar and Gowrish, Koushik Padavalli, "STATISTICAL FLOW-BASED TRAFFIC CLASSIFICATION FOR FULLY ENCRYPTED NETWORKS USING PURPOSE-BUILT LARGE LANGUAGE MODELS (LLMS)", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11302