A novel approach to detect IoT malware by system calls using Deep learning techniques

Recently, IoT devices or smart objects widely utilized in all kinds of fields such as medical, defense, automobile industry etc. Due to its intelligence and popularity, attacker seeks its help to launch malicious attack in high speed at low cost. In this regard, Researchers have turned their interest towards improving the security for the Internet of things devices. In this model, the malware were detected based on their behavior in terms of system calls sequence arise during its execution. The system calls of IoT malware are gathered using Strace tool in Ubuntu. The generated malicious system calls are preprocessed by n-gram techniques to retrieve required features. The extracted system calls were classified into two class i.e normal and malicious sequence using Recurrent neural network(RNN). The efficiency of this deep learning is tested using various performance metrics. The real time IoT malware samples were collected from IOTPOT honeypot which emulates different CPU architecture of IoT devices.