Machine Learning based Robust Techniques to Detect DDoS Attacks in WSN

Abstract

One of the most hazardous threats in the current environment is DDOs attacks. It is challenging to identify and defend against these attacks since they are becoming more sophisticated and more frequent every day. So it’s important to identify and prevent such type of attack before any impact. In this study, research on a new dataset of DDoS attacks which includes HTTP Flood attacks, and UDP Flood is conducted.SVM, KNN, and Random Forest among other machine learning algorithms for classification are used in this work to classify nodes as malicious or non-malicious nodes. Additionally, Fuzzy inference rules are applied to malicious nodes classified by machine learning algorithms(SVM, KNN, and Random Forest)accurately identify nodes as highly malicious, moderate malicious, or nonmalicious. Finally using all the processed information a decision is made to eliminate all the highly malicious nodes. Moreover, we have created a dataset namely AVV-DDos2286 which will be made publicly available for further studies on DDO attacks. Our proposed method uses a better hybrid approach combining machine learning algorithms with Fuzzy logic systems which outperform conventional DDos attack detection systems implemented using stand-alone machine learning algorithms. The results obtained has an accuracy of about 94% for classification.