用精简版大模型+专家路由，高效识别加密流量
MERLOT: A Distilled LLM-based Mixture-of-Experts Framework for Scalable Encrypted Traffic Classification
- 用教师学生框架压缩GPT-2，生成小而准的分类专家
- 在10个数据集上超越或媲美顶尖模型，资源消耗大幅降低
- 适合需要低延迟高精度加密流量识别的部署场景
我们提出MERLOT，一种基于混合专家（MoE）架构的轻量化大语言模型优化方案，专用于加密流量分类。通过教师-学生范式进行模型压缩，从GPT-2-base衍生出的紧凑模型在保持高分类准确率的同时显著降低计算开销。这些模型作为特定任务专家，在门控网络调度下动态激活。与生成式方法不同，本方法直接使用最终解码器标记，以上下文特征嵌入为输入完成分类。在10个数据集上的实验表明，该方法性能优于或媲美现有最先进模型，同时大幅降低资源需求，展现出卓越的效率与鲁棒性。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。