Research developments, trends and challenges on the rise of machine learning for detection and classification of malware

These days, malware is a major issue for anybody who uses the internet. Malicious software has evolved into a new breed called polymorphic malware, which is far more versatile than earlier waves. Since polymorphic malware is constantly altering its signature to avoid detection, it is infamously tough to identify. Several different machine learning algorithms were used to identify malicious malware and other security risks. Due to its high detection ratio, the most accurate approach was selected for system utilisation. The confusion matrix provided more insight into the system's performance since it tracked the amount of false positives and false negatives. We found that malicious computer traffic may be discovered through our research into malware analysis and detection utilising machine learning approaches, which could lead to improved network security. Naive Byes, SVM, RF, and J48 were the four algorithms used. The data showed that DT (99%), CNN (98.76%), and SVM (96.41%) were the top three classifiers in terms of detection accuracy. On a dedicated dataset, we tested DT, CNN, and SVM for malware identification using a tiny FPR. The results for DT, CNN, and SVM were 2.01%, 3.97%, and 4.63%, respectively. These results are significant because malicious software is becoming more common and more complex.