A hybrid deep learning intrusion detection model for fog computing environment

Abstract

Fog computing extends the concept of cloud computing by providing the services of computing, storage, and networking connectivity at the edge between data centers in cloud computing environments and end devices. Having the intelligence at the edge enables faster real-time decision-making and reduces the amount of data forwarded to the cloud. When enhanced by fog computing, the Internet of Things (IoT) brings low latency and improves real time and quality of service (QoS) in IoT applications of augmented reality, smart grids, smart vehi-cles, and healthcare. However, both cloud and fog computing environments are vulnerable to several kinds of attacks that can lead to unexpected loss. For exam-ple, a denial of service (DoS) attack can block authenticated users by rendering network resources unavailable and consuming network bandwidth unnecessarily. This paper proposes an intrusion classification model using a convolutional neural network (CNN) and Long Short-Term Memory networks (LSTM) to obtain the advantages of deep learning methods in order to accurately predict such attacks. The proposed integrated CNN with LSTM-based Fog Computing Intrusion Detection ICNN-FCID model is used for multi-class attack classification. Our proposed model is demonstrated using NSL-KDD, a benchmark dataset, and pro-vides attack detection accuracy of about 96.5%. Comparisons of the accuracy of our model with both traditional and other recent deep learning approaches show that our model is superior in performance. The ICNN-FCID model can be used in fog layer devices where network traffic is monitored and the attacks are detected. As a result, the cloud server and fog layer devices can be protected from malicious users and are always available in providing services to IoT devices.