Intrusion Detection Using End-to-End Memory Network
Chun Long, Gao Xiaoxian, Jing Zhao, Wei Wan, Hanji Shen, Peng Gao · 2017
Domain knowledge plays an important role in many intrusion detection systems (IDSs). Current IDSs have accumulated a considerable amount of knowledge on attacks. A number of hybrid models have been proposed to make use of domain knowledge into anomaly detection. The hybrid models consist of two learning phases: pre-learning of knowledge and classification based on it, which complicate the training process. In this paper, we present an approach which combines knowledge learning and classification into a single phase to simplify the training process. The model is an adaption of end-to-end memory network which is a neural network with an external memory. The prior knowledge is used to assist the classification of different types of attacks in the model. Extensive experimental results are conducted against NSL-KDD dataset to evaluate the proposed approach. The results demonstrate that the memory network has a good performance on intrusion detection.