Fog Computing

Once the system can estimate and discover resources, a third category appears having the purpose of allocating IoT applications in close proximity to the users. This technique, called resource allocation, utilizes the knowledge of available resources to map parts of the applications at different edge devices such that its requirements are met. There are two different perspectives of the allocation: (1) it represents the initial deployment to the edge of the network, deciding where to map the application; and (2) it serves as a migration technique by self-adapting when a node has failed. Moreover, one challenge arises when sharing resources between distributed edge devices, i.e. a close collaboration between nodes enforced by security and privacy is required. Solving this challenge creates the fourth category, i.e. resource sharing.

Finally, the last technique is called resource optimization and is obtained by combining the aforementioned resource management approaches. The main objective is to optimize the usage of available resources at the edge according to the IoT application constraints. Usually, the developer creates the QoS requirement of his application before deploying it to the edge.

Adopting the vision of fog and edge computing, more applications that today reside in the cloud are moved to the edge of the network. By deploying and connecting IoT devices, we can transform our homes in a more digitalized environment that adapts automatically, based on our behavior. However, with such benefits arise a set of privacy and security issues that we must address. For example, one can easily study the behavior of a family by simply accessing the generated data from sensors deployed in the house. Hence, ensuring data privacy and security remains a crucial factor in the evolution of edge and fog paradigms.

To evaluate the security and privacy enforced in systems based on fog and edge devices, the designer can use the confidentiality, integrity, and availability (CIA) triad model, representing the most critical characteristics of a system [29]. While any breach of the confidentiality and integrity components yields a data privacy issue, the availability component refers to the property of the nodes to share their resources when required. Since fog and edge represents an extension of the cloud, such systems inherit not only the computational resources but also the security and privacy challenges. Besides these challenges, due to the deployment of devices at the edge of the network more security challenges appear. Yi et al. identify the most important security issues of fog computing as authentication, access control, intrusion attack, and privacy [9].

Considering the dynamic structure of an IoT network, authentication is an important key feature of fog and edge computing and was identified. as the main security issue in fog computing [20]. The authentication serves as the connectivity mechanism that allows to securely accept new nodes into the IoT network. By providing means to identify each device and establish its credentials, a trust is created between the new added node and network. The current security solutions proposed for cloud computing may have to be updated for fog/edge computing to account for threats that do not exist in its controlled environment [21]. One solution to securely authenticate edge devices is presented in [30].

A comprehensive study of security threats for edge paradigms (i.e. fog and edge computing, and MEC, among others) was presented in [31], where the importance of security is motivated for the overall system and each individual component. An edge ecosystem consists of different edge nodes and communication components, ranging from wireless to sensors and Internet-connected mobile devices, distributed in a multilayer fog architecture. While each individual component has its own security issues, new different security challenges appear by combining and creating an edge ecosystem. By reviewing the scope and nature of potential security attacks, the authors propose a threat model that analyzes possible security risks (see Table 2.1).

For this model, the authors in [31] discover all important components of edge paradigms and describe all attacks that can occur against them. As depicted from Table 2.1., we can observe that five different targets i.e. network infrastructure, service infrastructure composed of edge data center and core infrastructure, virtualization infrastructure and user devices [31] are identified. The network infrastructure represents the various communication networks that connect edge devices which an adversary can attack using one of the following: denial of service (DoS), man-in-the-middle attacks, and rogue datacenter. An example of a man-in-the-middle attack on an IoT network is presented in [32]. On the one hand, an adversary could attack the service infrastructure, at the edge of the network, by using physical damage, rogue component privacy leakage, privilege escalation, and service or virtual machine (VM) manipulation. On the other hand, the core infrastructure is more secure being prone to attacks like rouge component, privacy leakage, and VM manipulation [31]. Finally, the virtualization infrastructure is exposed to attacks, such as DoS, privacy leakage, privilege escalation, service or VM migration, and misuse of resources; while user devices are susceptible to attacks like VM manipulation and injection of information.

Table 2.1 Threat model for fog and edge computing [21].

Privacy, defined as the protection of private data, ensures that a malicious adversary cannot obtain sensitive information while data is in transit [33]. At the moment, privacy is most vulnerable since the data of end users is sent directly to the cloud. From this point of view, edge and fog paradigms enforce privacy by moving the computation closer to the user. In doing so, data can be processed locally and the user can control what third parties are accessing his private data based on a defined role-based access control policy. However, some privacy challenges remain open, such as (i) the awareness of privacy in the community where, for example, almost 80% of WiFi user still use their default passwords for their routers and (ii) the lack of efficient tools for security and privacy for constrained devices [3].

The network management plays the most important role in both edge and fog paradigms since it represents the means of connecting all smart devices at the edge and ultimately providing available resources by deploying more nodes. Since the nature of an IoT network consists of heterogeneous devices, which are highly dispersed across large areas, an engaging task is to manage and maintain connectivity. Newly emerging technologies like software-defined networks (SDNs) and network function virtualization (NFV) are seen as a possible solution that may have a significant impact in implementing and maintaining the network increasing the scalability and reducing cost [19].