MERLOT: A Distilled LLM-based Mixture-of-Experts Framework for Scalable Encrypted Traffic Classification

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Yuxuan, Li, Rongpeng, Zhao, Zhifeng, Zhang, Honggang
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909397038399488
author Chen, Yuxuan
Li, Rongpeng
Zhao, Zhifeng
Zhang, Honggang
author_facet Chen, Yuxuan
Li, Rongpeng
Zhao, Zhifeng
Zhang, Honggang
contents We present MERLOT, a scalable mixture-of-expert (MoE) based refinement of distilled large language model optimized for encrypted traffic classification. By applying model distillation techniques in a teacher-student paradigm, compact models derived from GPT-2-base retain high classification accuracy while minimizing computational costs. These models function as specialized experts in an MoE architecture, dynamically assigned via a gating network. Unlike generation-based methods, our approach directly classifies encrypted traffic using the final decoder token with contextual feature embedding as input. Experiments on 10 datasets show superior or competitive performance over the state-of-the-art models while significantly reducing resource demands, underscoring its effectiveness and robustness.
format Preprint
id arxiv_https___arxiv_org_abs_2411_13004
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MERLOT: A Distilled LLM-based Mixture-of-Experts Framework for Scalable Encrypted Traffic Classification
Chen, Yuxuan
Li, Rongpeng
Zhao, Zhifeng
Zhang, Honggang
Machine Learning
Cryptography and Security
We present MERLOT, a scalable mixture-of-expert (MoE) based refinement of distilled large language model optimized for encrypted traffic classification. By applying model distillation techniques in a teacher-student paradigm, compact models derived from GPT-2-base retain high classification accuracy while minimizing computational costs. These models function as specialized experts in an MoE architecture, dynamically assigned via a gating network. Unlike generation-based methods, our approach directly classifies encrypted traffic using the final decoder token with contextual feature embedding as input. Experiments on 10 datasets show superior or competitive performance over the state-of-the-art models while significantly reducing resource demands, underscoring its effectiveness and robustness.
title MERLOT: A Distilled LLM-based Mixture-of-Experts Framework for Scalable Encrypted Traffic Classification
topic Machine Learning
Cryptography and Security
url https://arxiv.org/abs/2411.13004