(POSTER) Federated Learning Assisted Model for Android Malware Detection using Gannet Optimization Algorithm
Shikha Arya, Sateesh Kumar Peddoju · 2024
The prevalence of Android malware continues to rise, and we are more concerned about the privacy-preserving techniques employed in Android malware detection. In this paper, we leverage the efficacy of Federated Learning (FL) in handling such class imbalances while preserving data privacy in a distributed environment. In our experiment utilizing the CIC-Maldriod2020 dataset, known for its skewed distribution with scarce malware samples, we employ the enhanced Synthetic Minority Over-sampling Technique (SMOTE) to generate synthetic malware instances of the minority class. Further, we integrate feature optimization techniques based on the Gannet optimization algorithm to mitigate the impact of irrelevant features and enhance model performance. We apply the deep learning method of Artificial Neural Network(ANN) and Convolutional Neural Network(CNN) models to classify the samples into malware and benign. Our proposed work analyzes with a higher accuracy of 95.6% from CNN classification with optimization method compared to ANN. We also present the low error rate of MAE, MSE, and RMSE with optimization and without optimization methods using both CNN and ANN classification models in a federated environment.