Network Anomaly Attack Detection System Based on Incremental Learning Combined with SCV and SVM Algorithms

Network anomaly attack detection refers to analyzing network behavior and extracting features to predict and judge it. Since network anomaly attacks are a dynamic and uncertain behavior, their behavior is fuzzy and complex, so it is difficult for traditional data flow analysis methods to detect network anomaly attacks effectively. To improve detection accuracy, a network anomaly attack detection system based on incremental learning combined with a candidate support vector algorithm is proposed to optimize the support vector machine. Based on the historical data, the supporting candidate vector is proposed. This study adopts the retention strategy to classify the old samples. Combined with the Kuhntak condition, the new samples are screened, and the number of training samples in the incremental process is reduced to improve the efficiency of the detection model. The updated algorithm has a higher anomaly detection rate than the classical support vector algorithm. The normal false alarm rate of the model is 91.7%, and the abnormal detection rate is 3.5%. The classification effect of the final detection model is better than that of the classical support vector algorithm. This method has good accuracy and real-time performance in predicting and judging network anomaly attacks. © (2024), (International Journal of Network Security). All rights reserved.