Detection of <scp>DDoS</scp> Attack Using Taylor Elephant Herd Optimization ( <scp>TEH</scp> ) and Univariate Ensemble Feature ( <scp>UEF</scp> )

ABSTRACT The digital world has significantly evolved with advancements in internet services and cloud computing. While these improvements have enabled better data protection in cloud environments, they have also introduced new challenges due to the rapid development of various technologies. One of the greatest serious and ever‐growing threats is DDoS attacks, which can severely disrupt critical computer systems and render them ineffective. A novel approach is proposed to address this issue by integrating Taylor Elephant Herd Optimization (TEHO) with Univariate Ensemble Feature (UEF) selection. The bio‐inspired TEHO algorithm enhances intrusion network features, improving attack detection efficiency and solving complex network challenges. Meanwhile, UEF is employed to extract and refine relevant intrusion network features. The ensemble classification technique further strengthens detection accuracy by combining the results of four different classifiers through a majority voting process. This innovative method is crucial in fortifying the security of essential internet networks and systems against evolving cyber threats. The proposed TEHbUEF framework enhances DDoS attack detection through advanced feature extraction, optimization, and ensemble classification techniques. Initially, the UEF method identifies and selects the greatest significant features using Gain Ratio, Information Gain and Chi‐Square tests. These features are further refined using the Relief algorithm to reduce noise and focus on the most relevant data points. The TEHO algorithm then optimizes these selected features by iteratively adjusting their weights to improve their effectiveness within the network, significantly enhancing the system's ability to detect threats. Finally, an ensemble classification approach—combining Support Vector Machine (SVM), Naïve Bayes (NB), Logistic Regression, and Decision Trees (DT)—ensures robust and accurate detection through majority voting. The comprehensive nature of this approach leads to superior performance in accuracy, false detection rate, recall, precision, detection rate, and F‐measure, proving the effectiveness of TEHbUEF over existing methods. With an accuracy of 99.35%, recall of 99.36%, precision of 99.19%, detection rate of 98.47%, and an F‐measure of 99.28%, the TEHbUEF approach demonstrates outstanding performance. These results highlight its exceptional ability to detect DDoS attacks with greater accuracy and efficiency, surpassing existing techniques in both detection effectiveness and overall system performance.