A Lightweight malware detection technique based on hybrid fuzzy simulated annealing clustering in Android apps

Abstract

The growing complexity of cyber threats has shifted the focus from merely identifying threats to detecting their origins, resulting in stronger defenses against malware. Traditional detection techniques are often inadequate against increasingly sophisticated malware, prompting this research article to propose a new clustering method—fuzzy C-mean simulated annealing (FCMSA)—to enhance malware detection through machine learning. The FCMSA clustering technique improves performance by minimizing vulnerabilities, reducing outliers, and optimizing large datasets. The proposed technique selects high-quality clusters from Android app permissions and, using lightGBM, classifies Android malware. Experimental results show that the proposed FCMSA-GBM technique achieves superior accuracy (99.21%) and precision (99.70%) compared to other prevalent cluster-based Android malware detection techniques, while also lowering error rates and execution time.

Keywords

• Android malware detection
• Fuzzy c means clustering
• Simulated annealing
• Gradient boosting machine