STUDYING A SOLUTION FOR EARLY DETECTION OF DDOS ATTACKS BASED ON MACHINE LEARNING ALGORITHMS

Abstract

This paper focused on researching and proposing to build a system that acts as a sensor that can be installed anywhere on the network and performs online traffic classification. The proposed system used basic machine learning techniques for network anomaly detection and data dimensionality reduction techniques to remove features that are not significant in anomaly detection. The main goal of the proposed system was to reduce the computation time to help detect the attack early but still ensure the accuracy of anomaly detection. The obtained results showed that the model using the KNN algorithm combined with the feature extraction technique had relatively stable accuracy for all data sets (lowest is 99.15% on NSL-KDD set, highest is 99.73% in simulation dataset) with fast execution time (since the data is reduced in size, making the calculation faster).