Computer Software Malicious Intrusion Detection Method Based on CN and Dynamic and Static Behavior Analysis

As the information technology rapidly develops, the complexity and openness of the network environment have led to an increasing number of security threats such as malicious software and hacker intrusions, posing significant risks to personal privacy, corporate secrets, and national information security. To timely detect and prevent malicious attacks or abnormal behaviors, the optimization of capsule networks is studied, and a malicious software identification and detection method is proposed based on dynamic and static behavior analysis. This method converts malicious software code into images and utilizes capsule networks for feature learning and classification to achieve automated detection and classification of malicious software. The findings denote that the method performs well in detection accuracy, with a detection accuracy rate of 92.8% and a missing alarm rate of only 7.6%. Its detection time for a single sample is 0.8 ms. In terms of detecting different types of malware, the capsule algorithm has a detection accuracy of over 85.0%, with the highest being 96.2%. The research results show that compared with classic machine learning and deep learning baselines, this intrusion detection method has improved detection accuracy by 6.8%–12.4%, reduced false positive rate by 9.1%, and improved detection stability under confusion attacks by about 11.6%. It demonstrates significant quantitative advantages in both accuracy and robustness.