Optimizing Latency in Secure Reliable Transport (SRT) Protocol through Machine Learning for Video Streaming Applications

The increasing demand for live video streaming, especially using protocols like MPEG-TS and UDP/RTP, faces challenges of high latency and bandwidth overhead. Secure Reliable Transport (SRT) protocol, a superior open-source technology, optimizes streaming over unpredictable networks by managing packet loss, jitter, and bandwidth fluctuations, ensuring high-quality video. However, SRT struggles with latency issues due to packet delivery times, clock discrepancies, time drift, and Round-Trip Time (RTT) fluctuations. This paper introduces a machine learning approach to enhance streaming efficiency by predicting the optimal codec (MPEG4, H264, VP8) using K-means clustering to categorize codec groups and K-nearest neighbors algorithm for training the prediction model. This system significantly enhances bandwidth and reduces latency, jitter, and packet loss, achieving a testing accuracy of 98%. The results indicate a substantial improvement in latency optimization through machine learning in the SRT protocol, highlighting potential bandwidth and performance gains.