Hybrid Model for Ddos Assault Detection by Integrating Anomaly Detection and Behavioral Analysis Algorithms
Anil Suhag, Avneesh Kumar, Sudeept Singh Yadav · 2024
The paper proposes a novel hybrid model for detection of Distributed Denial of Service assaults by combining the outputs of behavioral analysis by Long Short Term Memory networks and anomaly detection by Random Forests at the decision level. The model employs the fusion technique of voting to integrate the predictions of the two models and finally, generates the final predictions for each instance based on the combined outputs. The research paper has evaluated the performance of existing anomaly detection algorithms like Isolation Forests, Random Forests, Long Short Term Memory networks and auto-encoders and existing behavioral analysis algorithms like Long Short Term Memory networks, Hidden Markov Model, Self Organizing Maps, Generative Adversarial Network, and Random Forests in terms of detection rate, accuracy, true positive rate, false positive rate and computational efficiency. The paper also demonstrates the effectiveness of integrated approach in detecting DDoS assaults with high accuracy and minimal false positives. The proposed model achieves detection and accuracy above 99% with False positive rate of 0.5%.