Research on unknown attacks detection based on ArtiAnomalyG algorithm

Distinguish accurately known and unknown attack is a key element to classification model in order to increase detection accuracy of unknown attacks. This paper proposes ArtiAnomalyG algorithm which expands sparse region and injects anomaly instances given the class label anomaly into the training data. Filtering experiment result shows that possibility of causing the collision with original instances is very small. RIPPER algorithm is chosen to induce classification model over expanded training data. Experiment results indicate that using ArtiAnomalyG algorithm could help classification model identify known and unknown attacks effectively.