In recent times, there has been a huge upsurge in malicious attacks despite sophisticated technologies in digital network data transmission. This research proposes an innovative method that utilizes the forward-propagation workflow of the convolutional neural network (CNN) algorithm to detect malicious information effectively. The performance comparison of this approach was accomplished using accuracy, precision, false-positive and false-negative rates with k-nearest neighbor (KNN) and support vector machine (SVM) algorithms. To detect malicious packets in the original dataset, an experiment was carried out using CNN's forward-propagation workflow method (N = 11) as well as the KNN and the SVM machine learning algorithms with a significant value of 0.005. The accuracy, precision, false-positive and false-negative rates were evaluated to detect malicious packets present in normal data packets. The mean performance measures of the proposed forward-propagation method of the CNN algorithm were evaluated using the Statistical Package for the Social Sciences (SPSS) tool. The results showed that the mean accuracy (98.84%) and mean precision (99.08%) of the proposed forward propagation of the CNN algorithm appeared to be higher than the mean accuracy (95.55%) and mean precision (95.97%) of the KNN algorithm, as well as the mean accuracy (94.43%) and mean precision (94.58%) of the SVM algorithm. Moreover, the false-positive rate (1.93%) and false-negative rate (3.49%) of the proposed method appeared to be significantly higher than the KNN algorithm's false-positive (4.04%) and false-negative (6.24%) as well as the SVM algorithm's false-positive (5.03%) and false-negative rate (7.21%). Hence, it can be concluded that the forward-propagation method of the CNN algorithm is better than the KNN and SVM algorithms at detecting malicious information.
Keywords: convolutional neural network; deep learning; k-nearest neighbor; machine learning; novel forward propagation; support vector machine.