Towards Adaptive and Scalable DDoS Detection in Distributed System using Tuned Hierarchical Machine Learning Techniques

A DDoS attack detection system is a software that constantly monitors the network or system activities in order to identify the anomalous patterns and detect the activities that are potentially malicious or harmful in nature. In the present era, cybercriminals provide a wide range of strategies to access confidential data. With the growth of the Internet, new threats emerge -mainly viruses and worms. Most, however, exploit vulnerabilities as are exposed in discussions regarding a variety of approaches such as password-cracking and interception of unencrypted data. Thus, security measures are effected to prevent systems from unauthorized access. Firewall technology is the most popular type of the defense mechanism to try to protect private networks from other types of attacks on the outsideIDS plays a significant role in network security, healthcare applications, credit card fraud detection, and insurance sectors. The main goal of research to be conducted within this study is a comparative analysis of DDoS detection, based on the different machine learning and deep learning models described above. Some machine learning approaches, including Random Forest (RF), SVM, and Gradient Boosting (GB), are used to develop IDS with realtime datasets such as DDoS attack detection set and the UNSW datasets for attacker networks. Gradient Boosting, which is an effective classification algorithm that performs complex interactions of multiple independent variables in relation to one or several categorical target classes, mostly is the choice behind this approach. In this approach, decision nodes are assigned adaptive weights, with greater weight assigned to the attributes that highly influence a given node, whereas less impactful features exert little influence on the prediction outcomes. The accuracy and performance evaluation metrics will be the basis for validation of the proposed system, implemented in Python.