Scalable Static analysis is progressively considered as the basis of Android malware detection, but the raw permission vectors are sparse, noisy and hard to interpret in the changing app ecosystems. In this paper, the authors of the research hypothesise a semantic permission-based detection pipeline to compress Android permissions into understandable capability sets and combine objective weighting with multi-criteria decision making (MCDM) to extract audit risk indicators. Using a conservative VirusTotal-based labeling policy to construct a balanced AndroZoo subset (38,062 apps) and test ablation baselines (becoming successively more inclusive of raw permission binary, semantic category statistics, objective weights (Entropy and CRITIC) and MCDM risk scores (TOPSIS/VIKOR). Using stratified holdout split, the optimal configuration (hybrid features with Random Forest) has an Accuracy of 83.16 and ROC-AUC of 91.73 (F1 = 83.25, MCC = 0.663), indicating that the inclusion of MCDM-derived risk signals can both increase the performance in terms of Discriminative and Increase the interpretability. The suggested design is aimed at large-scale screening processes whereby predictive and transparent risk cues should be present besides prediction. It has limitations: it depends on VirusTotal-style labels (subject to engine disagreement and temporal drift) and the existing evaluation is not yet external-dataset-generalization, temporal split validation, or robustness against obfuscation, and these are put as prioritized extensions to future work.