Article Preview
TopIntroduction
Fog computing, a term invented by CISCO (2015), has emerged to satisfy the enormous volume and variety of data resulting from IoT devices. It provides a complementary approach and decentralized computing infrastructure to solutions based exclusively on a cloud computing infrastructure. This is because transporting the data to distant and far-situated cloud servers consumes considerable bandwidth, which is not suitable for real-time analytics and latency-sensitive applications. Hence, fog computing represents the recent extension of cloud computing that introduces a set of nodes (computing, storage, and networking services) down to the fringes of the network to satisfy IoT applications constraints: low latency, location awareness, wide-spread geographical distribution, mobility support.
The fog computing paradigm has attracted considerable attention in academics and industry in short period of time. This emergent paradigm has the same characteristics of distributed computing, a computing form in which the data and the applications are distributed on several computers or systems, but connected and integrated by means of network services and interoperability standards. Fog computing architecture involves not only the interaction of software components, but also interaction between physical components (OpenFog, 2017), as well as associated virtual components (physical devices and parallel Cloud computing centres). These multi-layers collaboration requirements create a new set of security issues related to identity management, resource access management, and separation management concerns. In this context, several excellent models of fog computing architectures are available (OpenFog, 2017; CISCO, 2019; Farahani et al., 2018; Sohal et al., 2018; Zhao et al., 2019; and Iorgam et al., 2018).
Preliminary work of software engineering community consists in proposing architectural models for the description and the understanding of these innovative systems. Consequently, within the framework of systems engineering, the traditional development is no longer the desired solution to conceive fog systems. Existing fog computing architectures such as: (OpenFog, 2017; CISCO, 2019; etc.) differ from one reference to another but they all share three common components: “IoT”, “Fog computing” and “Cloud computing”. Each component has its own properties and complements the other ones to perform their tasks. The cloud includes data centres and ensures the storage and processing of large data. The fog represents an intermediate layer. It includes fog nodes capable of storage, calculation and networking. The IoT constituent combines the devices responsible for collecting data (sensors) as well as the actuators to execute the actions. IoT objects of different types send the data collected continuously to the nearest fog node, so that it achieves analysis and calculation mechanisms and then, decides which actuator it may be applied. Fog nodes are connected to the cloud only to transmit exceptional or large data.
Obviously, these architectures constitute an abstraction of the fog system; they represent a set of elements with a given intention. Thus, they give a simplified view of a part of the system hiding some difficulties. In practice, designing a secure fog system is too costly, and the software engineering community attempts to overcome this by integrating security considerations into the first activities of the fog systems life cycle. In our previous work (Bouhroum et al., 2019), the authors have also proposed a similar multi-views architecture for fog systems but in addition to the traditional layers (Cloud, Fog and IoT) recognized as necessary to manage a fog infrastructure, the authors have added two other layers for supporting some appropriate security services: “Security layer” and “Network access layer”. Thereby, several views can be considered: the structural view, the functional view and the security view.
In fact, all proposed architectures can discard the challenges due to the complexity of fog systems while offering a high abstraction and a clear and succinct separation between concerns, however, the actual difficulty behind fog computing modelling, given its distributed nature, is to provide a model being sufficiently expressive while seamlessly integrating and validating security considerations. At the heart of formal methods, constructing formal specifications of fog system is an effective approach and offers a higher-quality, more secure solution compared to conventional design methods; “formal methods are an effective approach to ensure systems reliability, by providing a high evaluation assurance level ‘EAL7’ “, according to Common Criteria (2012).