ANNIDS: intrusion detection system based on artificial neural network
Yanheng Liu, Daxin Tian, Aimin Wang · 2004
This paper describes a network intrusion detection system based on artificial neural network (ANNIDS). The advantage of neural network ensures that ANNIDS does not need expert knowledge and it can find unknown or novel intrusions. The key part of ANNIDS is an adaptive resonance theory neural network (ART). ANNIDS can be trained in real-time and in an unsupervised way. A weight hamming distance method is used in detection, which is simple and correct in finding anomalous behavior. A well-trained ANNIDS can monitor the network in real time. The experimental results show that ANNIDS performs best when vigilance parameter is 0.4 to 0.5 and intrusion threshold is 0.4. The false positive error is about 8%, the negative error is about 2%, and the total error is lower 10%.