Published Paper Details:

DDoS attacks are increasingly sophisticated, cybersecurity is a major concern for the safety of the contemporary world, making refined and customizable detection mechanisms a requirement. This study will employ 5 machine learning algorithms - Random Forest, Support Vector Machine (SVM), decision tree model, K-Nearest Neighbors (K-NN) and the logistic regression model in analyzing CICIDS 2017 dataset on which algorithms is best suited to detect the DDoS attack. It was found that important features such as flow duration, packet size, and protocols (type) and TCP flags were extracted and normalised to enhance the performance of the model. The model was assessed based on the accuracy, precision, recall and F1-score. The best performance obtained from all the testing algorithms models was presented by the Random Forest model, with an accuracy of 95.1 % of the results of malicious traffic detection, precision 94.3 %, overall 95.5 % round as well as 94.9 % F1-score, making the Random Forest model strong, reliable in the terms of any detections of the malicious traffic. The findings can emphasize on the efficiency of ensemble approach in minimising threats against the network and thus they can provide substantial reference of how intelligent cybersecurity solutions can be put into practice, in psuedo-real time scenarios Moreover, the paper also mentions the obstacles and limitations to the models like scalability weaknesses, overhead efficiency, and flexibility to new patterns of attacks, thereby indicating the future experimentation with resource-friendly and hybrid detection models.