A Cybersecurity Threat Detection Using Advanced Neural Network Methodologies

As the sophistication of cybersecurity threats continues to rise, colleges and universities are increasingly faced with threats to their digital assets. Traditional approaches to forecasting cybersecurity threats tend to fall short in accurately classifying specific threats, processing ultra-high-dimensional data, and responding suitably to patterns of attacks. To address these challenges, this study introduces a novel prediction model: BPNN-CPOA-CSP-UAC. The model incorporates a Backpropagation Neural Network (BPNN), a heavily used model that can learn complexities in data. To address parameter tuning deficiencies, the model employs the Crested Porcupine Optimization Algorithm (CPOA) to discover the best weights and biases for the neural network, significantly enhancing learning performance and predictive accuracy. Further performance improvements are achieved using the Harbor Seal Whiskers Optimization Algorithm (HSWOA) for feature selection. This eliminates duplicate information by choosing the most descriptive features and eliminating irrelevant information, thus improving the model's accuracy. Tanh-estimator normalization is also used during pre-processing to scale input features and thus ensure high-quality data during training. The model was trained and tested using two benchmark intrusion detection datasets, CIC-IDS-2017 and UNSW-NB15. Performance on testing is remarkable with precision and accuracy both at 99.99%, recall, and F1-score at 99.98%. The model also yielded a minimum false positive ratio of 0.0175%, a false negative ratio of 0.0165%, and a quick execution time of 90.5 milliseconds. These results suggest BPNN-CPOA-CSP-UAC as a practical and efficient academic institution cybersecurity threat prediction model. Since it can predict threats quickly and effectively, it is a strong defense tool against emerging cyber threats in educational institutions.