Classification of FPGA Based Network Traffic Using Machine Learning

Abstract

To successfully control infrastructure, the actual identification of internet activity is important. Including faster response times, classifiers automated segmentation strategies have demonstrated favourable performance. Throughout this document, utilizing regression analysis and decision tree features collection, the box and hydraulic model’s appropriateness is checked. Besides, the appropriate number of shipments found in a flow when obtaining flow-level characteristics is calculated by the tests conducted, showing that 46% of movement packages are a reasonable balance that guarantees high efficiency in the minimum time consumed. The tests’ findings show that randomized forests outperform other architectures with a peak precision of 35%. However, since software-driven classification algorithms do not fulfil the expected real-time specifications, we propose a statistical learning device based on the Field-Programmable Gate Array (FPGA) that uses an incredibly parallel architecture simplify the mechanism in this way. The proposed architecture achieves better performance, exceeding the recorded hardware-based classifier outputs using equivalent methods, which guarantees realistic driving management consistency in cluttered computer servers.