Affiliations: [a] Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Arau, Perlis, Malaysia
| [b] Faculty of Computing, Universiti Malaysia Pahang, Pekan, Pahang, Malaysia
Correspondence:
[*]
Corresponding author. Ahmad Firdaus, Faculty of Computing, Universiti Malaysia Pahang, 26600 Pekan, Pahang, Malaysia. E-mails: [email protected] and [email protected].
Abstract: Many smart mobile devices, including smartphones, smart televisions, smart watches, and smart vacuums, have been powered by Android devices. Therefore, mobile devices have become the prime target for malware attacks due to their rapid development and utilization. Many security practitioners have adopted different approaches to detect malware. However, its attacks continuously evolve and spread, and the number of attacks is still increasing. Hence, it is important to detect Android malware since it could expose a great threat to the users. However, in machine learning intelligence detection, too many insignificant features will decrease the percentage of the detection’s accuracy. Therefore, there is a need to discover the significant features in a minimal amount to assist with machine learning detection. Consequently, this study proposes the Pearson correlation coefficient (PMCC), a coefficient that measures the linear relationship between all features. Afterwards, this study adopts the heatmap method to visualize the PMCC value in the color of the heat version. For machine learning classification algorithms, we used a type of fuzzy logic called lattice reasoning. This experiment used real 3799 Android samples with 217 features and achieved the best accuracy rate of detection of more than 98% by using Unordered Fuzzy Rule Induction (FURIA).