This study addresses the growing attack of cyber-attacks on computer internet networks, in critical information infrastructure. The study attempts to improve detection in these networks by comparing three methods: Deep Belief Network (DBN), DBN with Isolation Forest, and DBN with Support Vector Machine. The quantitative methodology assesses the effectiveness and accuracy of various procedures in detecting abnormalities and provides numerical performance metrics. The results suggest that DBN alone is an excellent detection method for attacks, with good accuracy, precision, and recall. Furthermore, collaborative models that include DBN, Isolation Forest, and SVM show enhanced overall performance by exploiting the benefits of each method. This study has major implications for addressing security flaws and inefficiency in detection on internet networks, which is consistent with the problems raised earlier. The favorable findings of this study provide hope for the application of DBN technology, which will enable the strengthening of cybersecurity systems under legislation such as the Presidential Regulation on the Protection of Critical Information Infrastructure. The integration of DBN with other detection methods appears to be a promising strategy for improving security and contributing positively to national cyber defense.