Edge-Accelerated Analytics for Encrypted Network Traffic using Optimized Machine Learning

This paper addresses the growing challenge of analyzing encrypted network traffic for security purposes. It introduces ACETA, a system designed to accelerate encrypted traffic analytics (ETA) by leveraging x86 multicore platforms and software accelerators at the network edge. The research explores various data features and machine learning models and demonstrates that the integration of Intel's DAAL and OpenVINO libraries significantly reduces model training and inference times. The results highlight substantial performance improvements in ETA processing, offering a solution for timely and efficient network security monitoring.