Optimized Binary Classifier For Dos Detection Using Tabu Search Method

In the proposed system, an optimized binary classifier for the Denial-of-Services (DoS) attacks detection using the Tabu Search Method is created. The primary approach for DoS detection, in the current era, struggles with the problem of excessive false positive errors during its analysis of the data and also the wastage of time and cost during the computation, which is why it is difficult to get real-time threat prevention ready. The goal is to overcome these drawbacks, so the proposed system has a Tabu Search-based feature selection that improves the dataset by choosing only the most important features making classification more accurate and the computational time less. Actually, the binary classifier is trained using a hybrid machine learning approach. It is based on combining Support Vectors Machine (SVM) and Decision based Tree (DT) models. The Tabu based Search algorithm caters to an optimal feature selection by escaping local minima and frequently renewing search parameters to obtain the maximum classification performance. Additionally, the system undergoes an adaptive thresholding mechanism which not only makes it more resilient to variations of the attack patterns but also enhances its robustness by boosting the classifier threshold. The experimental Results show that the proposed system is capable of accomplishing a very high level of detection accuracy, rendering false positives irrelevant and so ensuring that the processing time is brought down compared to the conventional methods. This research highlights a scalable and effective method for the detection of DoS attacks in real-time in the context of current security frameworks.