Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a conference presentation introducing Trochilus, a novel framework that combines machine learning with programmable data plane processing to achieve high-throughput pattern matching for network security applications. Learn how this innovative approach addresses the challenges of maintaining performance and cost-effectiveness while handling growing network traffic and increasingly complex patterns in modern large-scale network systems. Discover the technical implementation that leverages a byte-level recurrent neural network (BRNN) to model complex patterns while preserving expert knowledge and enabling automated updates for sustained accuracy. Understand the semi-supervised knowledge distillation (SSKD) mechanism that converts the BRNN into a lightweight, data-plane-friendly soft multi-view forest (SMF) deployable as match-action tables. Examine the novel entry cluster algorithm that minimizes expensive TCAM requirements, making the solution scalable to large network environments. Review evaluation results demonstrating multi-Tbps throughput capabilities, support for various pattern sets, and high accuracy maintenance through automatic updates, presented by researchers from Peng Cheng Laboratory, Tsinghua Shenzhen International Graduate School, Cornell University, Hong Kong University of Science and Technology, and other leading institutions at USENIX ATC '25.
