Article Preview
Top1. Introduction
IoT becomes important technologies of everyday life. Internet of Things (IoT) consists of large number of networked physical devices such as computers, vehicles, digital devices, sensors etc (Ray, 2018). They send the data to cloud server and cloud provides response. There may be several kinds of time sensitive and time insensitive applications. Time sensitive applications may be suffered from the delayed response from the cloud server. In order to overcome the above problem, fog computing paradigm extends the cloud computing that provides services to end-users such as data, storage, computation and etc. (Yousefpour et al., 2019) (Puliafito et al., 2019). Fog layer acts as an intermediate layer between Cloud and IoT layer where heterogenous devices are present. Figure 1 shows the fog computing paradigm which consists of three layers such as IoT devices at bottom layer, fog computing as intermediate layer and finally cloud. Thus, time sensitive applications of health care, augmented reality and etc. get benefits by fog computing. Fog node can send immediate response to the end devices which gives low latency. Like the cloud environment, fog layer is also vulnerable to various attacks (Yi et al., 2015). So, it is necessary to have efficient intrusion detection system that contains set of mechanisms and tools. They monitor the computer system as well as network traffic. It should be able to analyse activities for detecting possible internal as well as external intrusions targeting the system in fog layer with the aim of protecting fog nodes as well as cloud. The IDS can be deployed on the fog layer to detect intrusive behaviour and malicious attacks such as denial-of service (DoS), Backdoors and so on. The IDSs are responsible for detection and prevention of attacks. In order to provide low latency through fog layer, it is really a challenge to implement efficient intrusion detection system. Thus, the Deep learning is a branch of machine learning that helps to implement intrusion detection system. Deep learning algorithms consists of multiple consecutive interlinked layers and perform their operations using layers (Zhang et al., 2018) (Alazab et al., 2019). The output of one layer will be the input of next layer.
Deep learning has proven to be effective in a variety of fields, including image and video recognition, audio processing, natural language processing and others. Machine learning approaches are used for implementing intrusion detection system. Shallow machine learning methods are ineffective in dealing the attack scenarios and difficult to deal security challenges due to the growth in new technologies and heavy traffic of Internet. The deep learning algorithms has proven to be effective in intrusion detection system like other fields. Deep networks can find the correlations among network traffic which contains both normal and anomalous records and can easily detect attacks. There is no need for human intervention. Deep learning approaches extract features and perform classification together.