Network Anomaly Detection Algorithm Based on Fusion of Artificial Fish Swarming and Isolation Forest
Chen Zhang, Wei Shen, Shengzhao Wang, Yue Wu · 2024
Currently, network security involves all aspects of human life privacy, etc., and anomalous data detection has been an important means of protection against network attacks. Aiming at the problem of unreasonable allocation of voting weights of different isolation trees in isolation forest, an isolation forest algorithm based on artificial fish swarming is proposed. The artificial fish swarming algorithm(AFSA) has a strong global search capability for the problem space solution, and in the stage of calculating the anomaly score, the artificial fish swarming algorithm is used to optimise the weight of the isolation trees in isolation forest, increase the weight of isolation trees with high detection accuracy, and reduce the weight of isolation trees with low detection accuracy, so as to ultimately improve the accuracy of the detection of anomalous data. In this paper ,we conduct comparative experiments with Probabilistic Generalization of Isolation Forest(PGIF) and Isolation Forest(IForest) on the ODDS anomaly detection dataset and network intrusion detection dataset. The experimental results show that the present algorithm improves the average AUC value by 6% and 5% on the ODDS dataset and NSL-KDD dataset, respectively.