Application of A Genetic Technique for Selection of the Best Features and Machine Learning for Malware Detection on Android

The open source Android platform, which is backed by Google, holds a significant share in the global marketplace. The prevalence of harmful programs has attracted the attention of hackers, who frequently conduct their activities by exploiting the widespread usage of the most commonly employed computer operating system globally. This study proposes an academic framework that advocates the utilization of a machine learning methodology to discern discriminatory traits. This strategy employs an evolutionary algorithm inspired by genetics to effectively detect instances of Android malware. Machine learning classifiers are taught using specific characteristics derived from genetic techniques. The effectiveness of these characteristics in detecting spyware is assessed before and after the process of information gathering. The result of the study offers evidence for the assumption that the evolutionary method is effective in optimizing the characteristics of the subset, resulting in a reduction of the characteristic dimensions to less than 50% of the amount of the initial attribute set. Machine learning classifiers demonstrate a persistent classification accuracy of more than 94% after choosing the features. This performance is achieved even while functioning on a reduced feature dimensions, thereby addressing the computational challenges associated with learning classifiers.