P. Zhang et al. / Future Generation Computer Systems 88 (2018) 16–27 17
Network equipment like switches and gateways is provided by
Cisco, Huawei, Ericsson, etc. The current research trends reflect the
tremendous potential of the Fog.
The Fog features with location awareness, low latency and edge
location [13]. It fits to a scenario where a huge number of heteroge-
neous ubiquitous and decentralized devices communicate, need to
cooperate, and perform storage and processing tasks [6]. Users can
visit their Fog anytime by using any device that can be connected
to the Fog network. The Fog has many applications in such areas
as smart city [14–16] and healthcare [17–20]. It can also provide
better Quality of Service (QoS) in terms of fast response and small
energy consumption [21,22].
The Fog uses network devices (named Fog nodes in this paper)
for latency-aware processing of data collected from Internet of
Thing (IoT) [23]. Fog nodes are denoted as heterogeneous com-
ponents deployed in an edge network in Fog environments. They
include gateways, routers, switchers, access points, base stations,
and specific Fog servers [24]. The Fog facilitates uniform and seam-
less resource management including computation, networking and
storage allocation [25]. Fog nodes are often the first set of proces-
sors that data encounter in IoT, and have the resources to imple-
ment a full hardware root of trust. This root of trust can be extended
to all the processes and applications running on them, and then to
the Cloud [26]. Without a hardware root of trust, various attack
scenarios can compromise the software infrastructures of the Fog,
allowing hackers to gain a foothold. The requirements of life safety-
critical systems mandate the sorts of security capabilities available
on the Fog [27]. Hence, new security and trust challenges emerge
with the rise of the Fog. The existing methods cannot be directly
applied to the Fog because of its mobility, heterogeneity, large-
scale geo-distribution [12]. This work reviews these concerns in
the Fog and the existing solutions. Differing from other survey
papers about Fog computing, this paper focuses on its security and
trust issues, especially in the region of the Fog.
The rest of this paper is organized as follows. Section 2 reveals
a Fog architecture as well as related security and trust issues.
Section 3 summarizes the related work to cope with security and
trust issues. Section 4 presents open research problems. Section 5
discusses the future work. Finally, Section 6 concludes this survey
paper.
2. Fog computing architecture
2.1. General architecture
Based on the modern computing architecture with three lay-
ers [11,21]: the Cloud, the Fog and the Edge, we provide a com-
prehensive fog architecture as shown in Fig. 1. Between the Cloud
and the Fog lies a core network to offer network services. From it
we can see that the Cloud lies at the upper core level and is far
away from edge devices. The Fog lies at the middle level and is
closer to edge devices than the Cloud. Each Fog node is connected
to the Cloud. Each edge device is connected to a Fog node [28]. In
addition, we can see that Fog nodes can be connected to each other.
Communications between Fog–Fog, Fog–Cloud, and Fog–Edge are
all bi-directional.
The Cloud: It includes high-performance servers and storage de-
vices for broadcasting, data warehousing and big data analysis [19].
It is the remote control and management center that can store large
data, and process highly complex but often non-urgent tasks. The
data is sent to the Cloud through high-speed wireless or wired
communications. The Cloud provides ultimate and global coverage.
As a repository, it provides data storage to meet users’ long-term
needs and intelligent data analysis.
The Fog: It consists of a network of interconnected Fog nodes
[19,24]. It provides geo-distributed, low latency and urgent com-
putation as well as location awareness. Each Fog node is a resource
Fig. 1. A comprehensive Fog architecture.
center for ephemeral storage. Its functions include network trans-
form, data collection, communications, data upload, data storage,
computation and management. Compared with the edge devices,
Fog nodes have more memory or storage ability for computing,
which makes it possible to process a significant amount of data
from edge devices. On the other hand, when needing a more
complex and longtime computation, the computation work should
be sent to the Cloud by Fog nodes through various available com-
munications technologies, e.g., 3G/4G/5G cellular networks and
WiFi. Fog nodes are bridges between Cloud and edge devices.
Fog nodes are independent and can be interconnected for col-
laboration. Management and collaborative procedures are applied
on Fog nodes to implement management and control. The col-
laboration among Fog nodes can be executed via remote or local
communications among them.
The Edge: It consists of several physical devices (edge devices)
enabled with their ubiquitous identification, sensing, and commu-
nication capacity [19], such as vehicles, machines and cell phones.
Each edge device is connected to one of the Fog nodes. Edge
devices have a large variety of sensors and local data. It is very
expensive and time-consuming to send all the data from terminal
edge devices to the Cloud through a network. Hence, by connecting
them to Fog nodes, one can deal with the urgent data but not
transfer from edge devices to the cloud immediately.
Some easily misunderstood concepts are discussed in the fol-
lowing.
Edge computing vs. the Fog: Edge computing is different from the
Fog in that the latter is a highly virtualized platform that provides
computation, storage, and networking services between end de-
vices and Cloud computing data centers [11]. Both of them need
to push intelligence and processing capabilities out of centralized
data centers and down closer to edge devices, such as IoT sensors,
relays, and motors. The key difference between them is where
intelligence and computing power are placed [12]. The Fog pushes
intelligence down to the local area network (LAN) level, processing
data in Fog nodes. While Edge computing pushes the intelligence,
processing power, and communication capabilities further down
to edge devices [29]. More details between them are discussed
in [24]. In [30], Endpoint computing is regarded as Edge computing.
Other similar concepts such as Cloudlets and Micro-data centers
are discussed in [31].
Wireless sensor networks (WSNs) vs. the Fog: WSNs are de-
signed to operate at very low power to extend battery life or
use energy harvesting to sustain themselves. Most of them face
the problems of small memory motes, low processing power, and