A Nature-Inspired Hyperparameter Optimization Approach for Android Malware Detection Using MLP and Cheetah Optimization Algorithm
Year: 2026
Type: article
Abstract: This research implemented a novel approach to malware detection by applying a COA-optimized Multilayer Perceptron (MLP). The research methodology used exploratory data analysis, data preprocessing via Standard-Scaler and SMOTE algorithms, and hyperparameter optimization using COA methods of search, sit-and-wait, attack, and retreat. Results of experiments demonstrated that the optimized version of MLP provided higher accuracy compared to its baseline variant; namely, from 97.52% to 98.07%. Moreover, the precision of the proposed method improved significantly, reaching 98% compared to 96.94% achieved before. Finally, COA-optimization allowed decreasing the number of trainable parameters by 32%; namely, from 3,297 to 2,241. Overall, the proposed study demonstrated promising results in terms of automated parameter selection, malware detection accuracy, and computational efficiency of the developed solution.
Institution: Federal University Wukari
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204106283