Transfer learning for raw network traffic detection

Traditional machine learning models used for network intrusion detection systems rely on vast amounts of traffic data with expertly engineered features. The abundance computational and expert resources at the enterprise level allow employment such models; however, these quickly dwindle in edge scenarios. As Internet Battlefield Things (IoBT) networks become common place tactical environments, there is a need improved distributed trained without resources. Transfer – which allows us to take information learned one domain apply it another provides way create distribute towards edge. Using neural networks, we demonstrate feasibility transfer using only raw computationally limited environments. Our results show that transferred one-dimensional convolutional model combined retrained random forest model, obtain over 96% accuracy 5000 training samples devices an time approximately 67 s.