Advancement of IoT Devices using LSTM for an Enhanced IoT Security
Nowadays, the evolve of unknown attacks have been prevented with an advancement of cyber security whereas there are several applications like smart cities and smart industries have evolved through Internet of Things (IoT) which perform one of the developing fields. The major challenge faced in the IoTs infrastructure is due to incremental of cyber-attacks fashions. The digital world development into an environment of physical gets accumulated with recent area attacks over traditional interne with current security threat. However, the main challenges faced in the physical IoTs connection is about implementation of distributed security mechanism to IoT devices resource constrain. It is essential that IoT devices can be automatically monitored and upgraded as firmware which consists of vulnerabilities like buffer overflows and it needed to be patched. In order to receive a replacement firmware, the devices are allowed to connect with cloud server automatically has been processed through firmware update. Anomaly and malicious behavior detection are the crucial concern which has become a priority in the area of intrusion detection. At present, the methods of intrusion detection can be generally based on single point detection and maintaining which can't be able identifies the attacking mode with the frequency of hidden attack. The familiar detection technology is recently using conventional Machine Learning (ML) algorithms for training the sample of intrusion to accomplish the intrusion detection models but these algorithms have demerits of poor detection speeds. However, the deep learning is one of the advanced technologies which extract behaviors from samples automatically. Hence, the intrusion detection accuracy is not high over traditional ML technology. Thus, the study introduced a hybrid algorithm of Elliptic Curve Cryptography- Long Short Term Memory (ECC-LSTM) model in which ECC perform as an Edge Node which exists over a private network and is built to perform as a connecting network client for an online cloud server is comparatively secured than of devices that may perform as a server itself. Eventually, the LSTM based on anomaly graph tool to detect Collective contextual Anomalies (CCA) has been detected from the samples and even detecting the malicious behavior from the extracted behavior chains which get evaluated with its accuracy, False Negative Rate (FNR) and False Positive Rate (FPR).