Delay Aware Recurrent-Convolutional Neural Network (R-CNN) for Ransomware Detection

Delay Aware Recurrent-Convolutional Neural Network (R-CNN) for Ransomware Detection

¹Tukur Adamu Muhammad, ²Muhammad Lamir Isah, ³Danlami Mohammed, ⁴Atiku Baba Shidawa, ⁵Mustapha Lawal Abdulrahman, ⁶Goteng Kuwunidi Job, ⁷Ismail Zaharaddeen Yakubu

^1,2,3,4,5,6 Department of Computer Science, Abubakar Tatari Ali Polytechnic, Bauchi

⁷SRM Institute of Science and Technology Kokalta, India

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - The adverse impact caused by ransomware on computing systems poses a major threat to everyday users and society in general. With continuous growth in ransomware, and newer malicious families emerging every month, the need for strong defensive methods increases every day. While expert-based systems are developed over time, this rate of growth in ransomware creates a need for self-evolving methods of defence that can learn from available data and improve over time. Deep learning methods, in particular, can provide this ability to improve learning with the increasing availability of data. This research proposes an alternative ransomware detection technique using deep learning, specifically Recurrent-Convolutional Neural Network (R-CNN), which will speed up the training time and extract high level features to give more accurate classification. Consequently, our model has improved the classification accuracy of the existing systems to 98.00%. The comparisons with other state-of-the art peer approaches have proven that our empirical model is promising.

Key Words:  Ransomware, R-CNN, Deep learning, Neural Network