IP3 2024 – AI based Network Management (Lot 15517)
Disclosed is a technique to determine an abnormal network traffic pattern using a bidirectional recurrent neural network (BRNN) to predict a pattern of very rarely observed traffic. Techniques are disclosed to collect traffic packets, extract particular field data from the traffic packets, and transform the extracted particular field data into a vector with a reduced dimension for each traffic packet. The system creates training data with the vector for each traffic packet and trains a traffic prediction model with the training data. The traffic prediction model predicts the next input traffic packet from an input traffic packet and determines whether the next input traffic packet is abnormal. Further disclosed that the system predicts with the trained traffic prediction model the frequency of abnormal traffic packets and outputs an abnormal traffic warning by comparing the predicted frequency with a threshold. The technology may be implemented in network traffic determination systems, network security systems, artificial intelligence-based network traffic management systems, etc.