Mobile Threat Detection System: A Deep Learning Approach

Even though there have been numerous approaches to tackle the security and privacy issues in mobile systems, there is still potential for improvement. Long lists of permissions are typically ignored by mobile users since they are confusing. Therefore, it is crucial to assess Android applications by determining whether they are safe or malicious and by ensuring that the likelihood of each permission request is well understood. In this paper, we propose an approach that combines dynamic analysis and deep learning techniques to tackle the security problems of Android devices. By applying deep learning algorithms on the famous Drebin dataset, we successfully recorded a record high accuracy rate of 99.20% with our DrebinDNN model in comparison to other deep learning models such as Recurrent Neural Networks(RNN), Radial Basis Function Networks (RBFN), and Self-Organizing Maps (SOM) and machine learning models including Random Forest (RF), Support Machine Vector (SVM), Decision Tree, and Naïve Bayes. With this encouraging performance, we believe that our approach holds the promise of constituting a reliable option to mitigate cyber threats targeting the Android operating system.