Rule Generation for Network Intrusion Detection Systems Based on Packets-To-Video Transformation

Modern intrusion detection systems utilize machine learning to identify network anomalies. Traditional static rules may not be sufficient to combat emerging attacks, making it critical to adopt a dynamic approach for keeping intrusion detection rules up-to-date. This study introduces an intelligent rule generator with a packet encoding method to represent packets into images, a vision model to encode the images, and a video captioning model, mapping image features to textual descriptions, thereby generating rules suitable for network intrusion detection systems. The results of our simulated data experiments show that our classification model has a higher accuracy than others and is capable of generating rules.