Malware Detection through Memory Dump Analysis by Enhanced Machine Learning Techniques

In this new era of the digital world, cybersecurity remains a critical concern, demanding frequent updates to counter sophisticated malware such as Trojan horses, Spyware and Ransomware. These threats have reached to high levels, outpacing traditional detection methods. To address this challenge effectively, we propose an innovative approach that combines Gradient Boosting and Random Forest classifiers in a ensemble method. Our study focuses mainly on analyzing the system's memory to identify malware, which often uses conventional detection mechanisms. The hybrid algorithm examines the patterns and behaviors within the system's volatile memory, significantly enhancing the accuracy and efficiency of malware detection. This approach involves collecting and analyzing memory information from diverse digital environments, combining both known and unknown malware scenarios. Initially, we explored the accuracy of four machine learning algorithms: Naive Bayes, Random Forest Classifier, Decision Tree, and Extreme Gradient Boosting. Among these, Random Forest and Extreme Gradient Boosting achieved an accuracy of 83.87% and 86.55% which was highest among the four machine learning models. Building on these results, we developed an ensemble method that blends Random Forest and Extreme Gradient Boosting classifiers. Remarkably, this blended ensemble method achieved the highest accuracy of 87.88%. In conclusion, our proposed approach not only achieves a balanced accuracy but also computational efficiency in malware detection. These findings represent a significant advancement in memory-based malware detection techniques against the evolving cyber threats.