INTRUSION DETECTION SYSTEM BASED ON SELF-ORGANIZATION MAP AND ON LEARNING VECTOR QUANTIZATION

Traditional intrusion detection methods lack extensibility and adaptability in face of unknown attack types. At the same time, current neural networks algorithms are generally based on supervised learning, and need labeled data for training first. Self-organizing map algorithm is used to detect novel attack in some papers[2,4,5].In this paper, self-organization map is adopted to classify the input data produced by detectors, followed by a updated learning vector quantization–LVQ3. Experiment results indicate that this algorithm is more effective for intrusion detection than traditional intrusion detections or the algorithm that only utilizes SOM.