Fuzzy Membership Grasshopper Optimization Algorithm (FMGOA) Based Feature Selection and Mean Weight Deep Belief Network (MWDBN) Classifier with Fusion Approach for Android Malware Detection (AMD)

Android applications have been an obvious advancement recently, making them one of the technological domains is advancing and successful the fastest. The necessity for active research efforts are presented to conflict these dangerous programs which develops vital as malware gets more and more capable of insightful these applications. In order to enhance Android Malware Detection (AMD), machine learning is becoming more and more popular. In this paper, feature selection and classification fusion strategy for AMD. Firstly, the dataset is gathered from samples of Android apps. Secondly, Fuzzy Membership Grasshopper Optimization Algorithm (FMGOA) is introduced to choose the most important features. FMGOA approach imitates the biological behaviour of grasshopper swarms searching for best selection of features with their accuracy. Thirdly, a Stacked Ensemble Classifier Fusion (SECF) is introduced based on a multilevel architecture-based approach. It enables the efficient merging of machine learning algorithms including J48, Reduced Error Pruning Tree (REPTree), Voted Perceptron, and Mean Weight Deep Belief Network (MWDBN).Two ranking-based algorithms—Ranked Aggregate of Average Accuracy and Class Differential (RACD) and Ranked Aggregate of Per Class (RAPC)—have been presented that enable classifier fusion for stacking. Finally, Precision (Pre), Recall (Rec), F-measure (FM), and Weighted F-measure (WFM) has been used to evaluate the results of classifiers.