Abstract:
The problem of Android malware detection in imperatively demanding conditions (i.e., an increasing number of applications are being developed at a fast pace as well as ever-growing threats) is investigated. In this article, we suggest a machine learning approach to the problem including data acquisition, pre-processing, feature engineering methods and models. We can see that Bagging Classifier achieves Top Performance metrics bagging the same with Precision, Recall and F1-score of 0.84, thus leading to Overall Accuracy being (0.84), though Random Forest astonishing too having metrics as 0.81 Experimental evaluation against existing techniques shows that our ensembles can detect intricate malware patterns, robustly addressing the limitations of metadata-based detection strategies. We also integrate Explainable AI methods to improve the interpretability and understandability of the decisions made by our model, making it easier for trust in cybersecurity implementations. The work underscores the necessity for evolving learning schemes to encompass new threats and future investigations aiming to exploit deep learning techniques, crossplatform detection strategies, and implications regarding user privacy and ethical concerns. This work gives a wealthier gainful binary to envision portable security look into field and in malware detection and avoidance specifically.
