ZERO-DAY ATTACK DETECTION MODEL FOR NETWORK INTRUSION USING THE CUCKOO SEARCH ALGORITHM AND MACHINE LEARNING CLASSIFIERS

Cybersecurity threats, especially zero-day attacks, present a major challenge to modern digital infrastructures because they exploit previously unknown vulnerabilities before security patches are released. Traditional intrusion detection systems (IDSs) often fail to detect these novel threats because they rely on predefined rules and known signatures. This study presents a zero-day attack detection model for network intrusion that combines the Cuckoo Search Algorithm (CSA) for feature optimization with five supervised machine-learning classifiers Artificial Neural Network (ANN), Decision Tree (C4.5), K-Nearest Neighbor (KNN), Naïve Bayes (NB), and Support Vector Machine (SVM). The publicly available TII-SSRC-23 and CIC-DDoS2019 datasets were used to train and evaluate the model, with data acquisition, preprocessing, CSA-based feature selection, classification, and performance evaluation implemented in Python. Model performance was assessed using accuracy, precision, recall, F1-score, kappa statistic, and time taken to build the model. Results show that CSA-based feature selection improved detection accuracy for most classifiers on the TII-SSRC-23 dataset, with ANN consistently achieving the best performance (99.03% accuracy, 0.9903 precision, recall and F1-score) while Naïve Bayes recorded the lowest performance overall. The developed model provides a scalable and efficient framework for enhancing network security against emerging zero-day exploits, and offers a benchmarked basis for future real-time and deep-learning extensions.