A Distributed Access Control Scheme based on Risk and Trust for

Fog-cloud Environments

Wided Ben Daoud

, Amel Meddeb-Makhlouf

, Faouzi Zarai

, Mohammad. S. Obaidat

and Kuei-Fang Hsiao

NTS’Com Research Unit, ENET’COM, University of Sfax, Sfax, Tunisia

Department of ECE, Nazarbayev University, Astana, Kazakhstan, University of Jordan, Amman, Jordan

Ming-Chuan University, Taiwan

Keywords: Fog Computing, Access Control, Risk, Trust, Resource Manager, Monitoring.

Abstract: Fog computing is a technology, which benefits from both Cloud Computing and Internet of Things paradigms.

Instead of centralizing the information stored in the cloud, the idea in the fog environment is to use devices

located at the edge of the network to prevent congestion during data exchanges. Since its development, the

fog computing was integrated in numerous applications, including, connected vehicles (v2x), and smart cities,

among others. Furthermore, the fog environment is highly variable within the resources as it varies

dynamically. Therefore, a careful management of resources is essential for maximizing the efficiency of this

distributed environment. That is why, the access control to resources is one of the challenging issues to be

taken into account to enhance security in fog environments. In this paper, we develop a novel distributed fog-

cloud computing mechanism for access control based on trust and risk estimation. We evaluate the

performance of the novel approach through appropriate simulations, which improves system performance by

minimizing the time spent to have permissions to access services and optimizing the overall system resource

utilization.

1 INTRODUCTION

There are recent technologies that are increasingly

making our lives more comfortable and more

convenient. Cloud computing is a good example on

these. It simplifies the storage and the management of

data with higher efficiency and lower cost. Another

new technology is the Internet of Things (IoT),

introducing and changing our daily lives is the

Internet of Things (IoT) (Almutairi et al., 2012), (dos

Santos et al., 2016), which connects physical objects

and enables these objects to collect and exchange

data.

Fog computing combines the benefit of both the

cloud computing and the Internet of Things. In brief,

the concept of fog computing is an extension of the

paradigm of cloud computing to the edge of the

network. A scalable and elastic fog computing

environment is intrinsically very dynamic. Thus, the

major advantage of using this new technology lies in

its capacity to not overload the information network.

Consequently, users can store their data online in

remote servers and access them from anywhere (dos

Santos et al., 2016). This makes the problem of

providing adequate access control in fog even more

difficult. In addition, as fog computing offers new

opportunities to exchange personal data between

different fog nodes and cloud service providers,

where security and privacy experience critical issues

in the expansion of any sensitive information. Given

that this paradigm is an open platform and the

interaction between participants among themselves in

a fog environment is high, this can cause various

security concerns. Clearly, there is a need to protect

the privacy of data from malicious attacks. Therefore,

we propose a novel distributed model of access

control to improve the existing mechanisms and,

especially to facilitate traffic management in fog

computing environment.

In this paper, we examine in more details the

access control mechanism for fog-cloud

environments. Our goal is to propose a

comprehensive model to control user access to the fog

computing environment. The purpose behind our

work is to ensure not only the privacy and the safety

of shared information, but also to reduce the

296

Daoud, W., Meddeb-Makhlouf, A., Zarai, F., Obaidat, M. and Hsiao, K-F.

A Distributed Access Control Scheme based on Risk and Trust for Fog-cloud Environments.

DOI: 10.5220/0006848502960302

In Proceedings of the 15th International Joint Conference on e-Business and Telecommunications (ICETE 2018) - Volume 1: DCNET, ICE-B, OPTICS, SIGMAP and WINSYS, pages 296-302

ISBN: 978-989-758-319-3

complexity of administration and management of

security mechanisms.

1.1 Need for Distributed Access

Control in the Fog-cloud

In a distributed environment such as fog computing,

access control issues pose serious challenges to the

widespread adoption of cloud-based IoT services.

Thus, the short distance between the user and the

information also offers greater efficiency and requires

more security. The fog computing is a dynamic

paradigm, which makes resources distributed and

pervasive. Therefore, we require a judicious

administration of resources in order to enhance the

effectiveness of the fog.

More importantly, the cloud-fog cooperation will

produce an enormous potential. Hence, to decrease

heavy computational and communication overhead

on service providers or data owners, we propose a

distributed architecture of access control.

1.2 Brief Description of Fog

Computing

The Fog Computing model is designed as an

extension of the cloud. The term "fog" was initially

proposed by Cisco to designate the need for a

supporting platform capable of ensuring the

requirements of IoT (Huang et al., 2017), (Shirazi et

al., 2017). Fog computing is described as a highly

virtualized platform that delivers storage, computing

and networking services between the terminals and

traditional cloud computing data centers, at the edge

of the network (Figure. 1). Thus, fog computing

supports cloud computing to be more evolving and

scalable (Hu et al., 2017).

Figure 1: The concept of Fog computing.

This paradigm is generally deployed in

applications requiring low latency. Actually, cloud

can easily host many fog applications. Table 1

presents the fog computing characteristics.

Table 1: Cloud and Fog characteristics requirements.

Requirements Cloud Fog

Latency High Low

Distance Through Interne

Limited Hop Coun

Bandwidth More Demand Less Demand

Mobility Limited Supporte

In fog computing, privacy of data is a requirement

that must be met when analyzing sensitive

information. In support of this requirement, optimal

and secured access control processes should be

considered. Besides, devices, as well as the service

providers are heterogeneous and distributed, which

may increase problems in managing resources and the

user’s access to these resources.

Subsequently, the application of security in fog

environments is a major challenge. The goal of access

control for services is to maximize the efficiency of

resource utilization, and satisfy the users’ security

and privacy needs, meanwhile, take full advantage of

the profit of both fog providers and cloud service

providers (Shirazi et al., 2017), (Dastjerdi and Buyya,

2016).

The rest of this paper is organized as follows.

Section 2 describes a trust-risk aware approach for

security improving access control scheme. Then, an

experimental evaluation of the proposed solution in

terms of performance will be discussed in Section 3.

Section 4 concludes the paper and provides

concluding remarks.

2 OVERVIEW OF THE

PROPOSED SOLUTION

To manage the scalability problem in terms of users

and resources, we propose a distributed access control

architecture in fog-cloud computing. Each fog node

or cloud server applies independent access control

policies, and it is necessary to have mutual access to

resources between domains, which implicates a

corresponding access control mechanism to manage

the interoperability inter-domain (Almutairi et al.,

2012). The proposed distributed architecture, which

is summarized in Figure. 2, processes and integrates

the risk and the trust concepts. Then, it is constructed

using three types of components, as shown in Figure.

4: Resources Manager (RM), Distributed Access

A Distributed Access Control Scheme based on Risk and Trust for Fog-cloud Environments

297

Control Administrator (ACA) and Service Level

Agreement (SLA).



Figure 2: The main strategy proposed for access control.

2.1 Risk and Trust based Access

Control Definitions

2.1.1 Risk

Risk estimation is one of the basis of the proposed

technique. It is defined by the likelihood of a

dangerous and harmful circumstance and its

consequences if it happens. All research has the

ultimate objective to reduce the possibility of

occurrence of a certain risk through the application

and the implementation of mechanisms of control and

policies in the system. Therefore, the techniques of

risk estimation are required to ensure an automatic

prevention and protection from insider attacks.

Currently, there are risk factors, which represent

all possible outcomes that can be found about a

costumer. The risk value is calculated after finding the

probability of each event, which is then multiplied by

a weight attributed by experts to each metric. These

factors are evaluated for every access request and

aggregated to achieve a measure of the total security

risk. Details about all the metrics can be found in

(Dastjerdi and Buyya, 2016).

2.1.2 Trust

Trust is another crucial concept in our approach. Trust

is defined as a personal, subjective “expectation a

manager has about another's future behavior based on

the history of their encounters” (Dastjerdi and Buyya,

2016), (dos Santos et al., 2016). The trust generally

depends on the context in which the interaction

between units takes place.

2.2 Distributed Access Control

Architecture

The distributed access‐control architecture combines

the following components: RM, ACA, and the SLA

established between fog nodes and the cloud service

providers (CSPs).

2.2.1 Access Control Administrator (ACA)

a. Components of ACA

This module is used at each service layer of the CSP

and fog nodes to ensure the access‐control at the

corresponding layer. Figure 3 shows the proposed

model, which contains the following components:

 Policy Enforcement Point (PEP): This block is

in charge of intercepting all the access requests of

users. It ensures that the resources of the system

can be accessed only if the policy authorizes it. In

fact, the PEP extracts the authentication

credentials, the user attributes and the context

information from the access request.

 Policy Decision Point (PDP): It evaluates all

necessary information/policies, according to the

trust or the risk estimation, and the context. It

requests behavior history of user from Monitoring

System-Database (MS-DB). Then, it returns the

grant or rejects decision to the PEP, which applies

and enforces the decision.

 Risk Evaluator: Each access request and each

permission is assigned a risk value within a

particular context.

The proposed architecture provides a risk-based

access control to enable a dynamic access to the

available resources. The calculation of risk

determines whether an event can cause a system to be

destroyed and its assessment is useful for making

decisions to avoid such damages.

 Trust Evaluator: The Trust evaluator uses the

monitored information to identify the context, and

calculate the trust value of each user.

If the trust value of a user decreases while a user has

a session open, the RM sends a notification to the

PDP, which re-evaluates whether the privileges that

the user is exercising should be revoked. In this way,

the system is able to deny access to misbehaving users

before they can perform extensive damage to the

system.

 Monitoring System: Each fog node and cloud

server are supervised by a monitoring system. The

monitoring is a process of continuous observation

that reports any violation of the fog-CSP. Users

WINSYS 2018 - International Conference on Wireless Networks and Mobile Systems

298

have to achieve their assigned obligations and

their final status reported by the monitoring

system and forwarded to the MS-DB.

The final report stored in the MS-DB will be used to

calculate the risk and the trust values. Besides, a

monitoring function is also developed to check if the

obligations are fulfilled or also to capture any

unsuccessful decision-making, and then, deactivate

such access permission when detecting suspicious

activities.

 Policy Information Point (PIP): The policy of

the system is stored in the PIP. 

 Trust DB: The trust values are stored in the

Trust Database.

b. ACA Procedure

Access requests of users, which include the

requesting subject and the resource requested, are

given to the PEP. The latter forwards the request to

the PDP. The PDP takes the final decision concerning

the access request by evaluating the risk that can

engender the user and the trust about the user’s

behaviors. This is done by considering the attributes

and the credentials of the requesting user. The PEP

receives the decision from the PDP and then it either

permits or denies the request. If it is allowed, it is

forwarded to the RM of the SP for the deployment of

requested resources.

The PDP retrieves the risk and the trust value of

the user and determines whether the user is

trustworthy or not. When the user is trusted enough to

complete successfully, the access is granted and the

Monitoring System will supervise it. In case the user

is not trustworthy enough to fulfil one or more of the

obligations that would be assigned to him, the system

denies the access request.

2.2.2 Resource Manager (RM)

This module handles the resource needs of clients.

This is a part of the access control mechanism of each

CSP and fog nodes. Moreover, this module manages

the resource requirements of each fog node and cloud

service.

The RM is responsible for the deployment of

resources. In fact, the requested resources are

managed directly by this module if they are available

locally. However, if the requested resources are

stored in foreign services (remote data center), the

RM contacts the ACA of the remote entity/node to

provide the requested remote resources without

repeating the access control mechanism. This

procedure is taken owing to SLA established between

the CSPs and fog nodes.

2.2.3 Resources Mapping

The access request received from the user contains

the amount of required resources. After verification

of the identity and making access decision, the PDP

assesses if the services corresponds to a local fog

node (the resources requested are locally hosted). If

the service does not exist locally, it is assumed that

this request needs services mapping by a relevant

SLA. The ACA party invokes the SLA, if the

requested resources are stored on a remote CSP.

Figure 3: Overview of the access control administration (ACA).

A Distributed Access Control Scheme based on Risk and Trust for Fog-cloud Environments

299

Figure 4: Distributed authorization process in the multi_fog_cloud.

2.2.4 Service Level Agreement (SLA)

The SLA is a contract by which a CSP undertakes to

provide a set of services to one or more customers

(Almutairi et al., 2012).

In the fog-cloud environment, SLA ensures

certain levels of security in the storage and

management of personal data. It is then necessary to

define very precisely different quality indicators that

can be measured, analyzed and checked regularly.

SLAs ensure access to remote resources while the

resource manager is responsible for monitoring and

checking deployed and demanded resources. To

allow and facilitate exchange of information among

the architecture entities, an SLA performs resource

mapping. For this purpose and in order to prevent

side-channel attacks, SLA implements separation of

constraints for resource sharing. In the proposed

approach, the ACA chooses whether to invoke the

SLA or not depending on the location of requested

resources (local or remote).

2.2.5 Distributed Authorization Process

As shown in Figure. 4, each fog node and CSP has

both the ACA module and the RM module to deal

with the access control and resource management in

order to address the mobility of users.

When a user requests a service or local resource,

the local ACA of the SP receives this request. If the

ACA allows access to this user, it forwards the

request to the local RM in order to have the requested

resources. Otherwise, if the required resources are

hosted in a remote cloud or remote fog node, the local

RM contacts the ACA of the foreign fog/cloud as the

appropriate SLA is already established between them.

After the receipt of the allowed access request, the

ACA of the remote cloud or fog node forwards the

request (since this request is permitted, the system

does not repeat the procedure of the access control

from the beginning) to the RM in order to assign the

requested resources. Finally, the RM saves the

identity of this user and monitors its behaviors, then

stores these in the MS-DB.

3 IMPLEMENTATION AND

RESULTS

In this paper, distributed access control architecture is

proposed. In this experiment, we develop a script,

which contains the activities process of our model.

The script is implemented under Linux/Ubuntu. We

deploy OpenStack to maintain the experiment.

OpenStack is a software allowing the construction of

private and public cloud. In addition, OpenStack is a

community that is designed to help organizations

implementing a server or virtual storage system.

For this work, several simulation experiments are

provided for the server service provider as well as for

the distant instance (Table 2).

Table 2: Real and Virtual Machine Condition of

Simulation.

Condition Openstack server Remote instance

RAM 1 Go 2048 Mo

OS 14.04 LTS 12.04 LTS

CPU 2 1

The experiment measures the estimated time to

reach an access decision using the evaluation of trust

and risk. When the user requests access to

openstack/cloud, the script will be executed. Then,

depending on the available resources, the system will

calculate the risk and thus, it has an access decision.

To calculate the risk value, we choose a limited

number of risk factors to make our experience. These

factors are evaluated for every access request and

aggregated to achieve a measure of the total security

risk. There are some factors that represent risks

associated with the person or the application

requesting access to a resource. Another ones are

related to the risks associated with components in the

path between requester and resource, such as the type

WINSYS 2018 - International Conference on Wireless Networks and Mobile Systems

300

of machines, and risks associated with the situation

surrounding the request. Additionally, there are the

heuristics, which represent the risk associated to

previous similar requests, such as known violations.

Therefore, equation (1) shown below presents the

total risk:

AccessRisk

∑















(1)

where Σα

= 1, α

: is the weight, m

is the used metric

and n: is the number of metrics.

Algorithm 1 resumes the developed process.

Algorithm 1: Algorithm for the proposed access control

scheme.

Our script contains risk calculation policies. At

each access of such user, this script will be executed

and we calculate the elapsed time to have a final

decision about access. In this work, we choose 7

metrics to calculate the risk; each used metric has a

weight of 1/7.

The risk factors used are:

- User_role: administrator, simple user …

- Previous violations: yes/ no.

- Machine type: Server, Pc, Mobile…

- Network: wireless/ wired

- Connection type

- Rank

- Distance between user and cloud data center.

Next, we compute the risk value based on the

previous factors like in (2). These values in (2) are

just an illustration and different systems can use

different values.

(2)

We simulate our scheme several times successively

and afterwards (7 runs), we averaged the results

obtained from each simulation run.

Figure 5: Average of time spent by number of user's access

to local/ remote resources.

Figure 6: Comparison between distributed non-distributed

models of access to remote resources.

From the results obtained in Figure. 5, we can see

that our proposed scheme does not introduce an

important time-based overhead. As our model is very

distributed, when the number of users increases, the

time spent to manage all requests remains short.

Furthermore, by comparing the time spent to access

local resources and the time spent to access remote

cloud server, we can note that the differences are not

very high since our access control model is

distributed, which enable to function faster than

others models.

Therefore, the delay elapsed to handle all access

requests is not large. Owing to the fact that our model

is associated with the monitoring function, when the

user requests service and it is not available in this fog

node, he/she will not repeat all the procedure by

searching again in other neighboring fog nodes. By

using our model in fog, the allocation of resources is

realized with a short delay due to the distributed

A Distributed Access Control Scheme based on Risk and Trust for Fog-cloud Environments

301

nature of fog. In fact, for higher number of dispersed

services, the allocation results in a shorter delay.

Actually, if the user is trustable, he will have his

requested services, whether this latter is either locally

in fog nodes or far in foreign cloud data center.

Hence, compared to other traditional solution like in

(Dastjerdi and Buyya, 2016) the time spent for access

to remote resources is decreasing. The strong point

of our model is its ability to not overload the network

and this by setting up an architecture of access control

and monitoring taking place in a short time and

indicating that it is fast and efficient. Furthermore, we

can say that our model is secure. Thus, it is based on

the calculation of risk and trust. Besides, the

monitoring system is present to supervise and

discover if there are violations or malicious activities.

It intervenes to protect the system against destruction.

In the concept of the proposed solution, if the

requested resources are located in foreign cloud

services, the system is able to bring these services

without repeating the access control procedure from

the beginning. In fact, due to the monitoring system

and the resource manager, the time spent will

decrease towards half and that is a good result, which

proves that the proposed scheme is effective.

Moreover, the experimental results show higher

performance when using a distributed architecture for

controlling the access to the system.

4 CONCLUSIONS

The main characteristic of fog computing is its

effective management of resources. The distributed

nature of this environment and the deployment of an

access control strategy bring many challenges such as

deciding on the extension of collaborative work and

the limit of resources sharing. Therefore, it is

necessary to secure access to these resources. Hence,

there is a need to develop a new distributed trust-

based access control models that are adaptable to fog

computing. In fact, to decrease heavy computational

and communication overhead on service providers or

data owners, we propose a dynamic and distributed

access control strategy for fog computing based on

risk and trust evaluation in order to improve the

efficiency of fog resources’ deployment and satisfy

the users’ security requirements.

We began by describing the fog paradigm and

then identifying the need for a distributed strategy for

controlling access to fog-cloud systems. Simulation

was performed with an OpenStack platform in Linux.

Simulation results show that our proposed distributed

scheme can provide secure and efficient access

control management for users and then improve the

utilization of fog-cloud services.

REFERENCES

Almutairi, A., Sarfraz, M., Basalamah, S., 2012. A

distributed access control architecture for cloud

computing. IEEE software, vol. 29, no 2, p. 36-44.

Dos Santos D. R., Roberto M., Gustavo, R.S., Carla M. W.,

Carlos B. W., 2016. A Framework and Risk

Assessment Approaches for Risk-based Access Control

in the Cloud, Journal of Network and Computer

Applications, vol. 74, p. 86-97.

Dastjerdi, A. V., Buyya, R., 2016. Fog computing: Helping

the Internet of Things realize its potential. Computer,

vol. 49, no 8, p. 112-116.

Huang, Q., Yang, Y., Wang, L., 2017. Secure Data Access

Control With Ciphertext Update and Computation

Outsourcing in Fog Computing for Internet of

Things. IEEE Access, vol. 5, p. 12941-12950.

Hu P., Dhelim S., Ning, H., 2017. Survey on Fog

Computing: Architecture, Key Technologies,

Applications and Open Issues. Journal of Network and

Computer Applications, vol. 98, p. 27-42.

Shirazi, S. N., Ul H., Gouglidis, A., Farshad, A.,

2017. Review and Analysis of Mobile Edge Computing

and Fog from a Security and Resilience

Perspective. IEEE Journal on Selected Areas in

Communications, Issue: 99.

Xiao, M., Zhou, J., Liu, X., 2017. A Hybrid Scheme for

Fine-Grained Search and Access Authorization in Fog

Computing Environment. Sensors, vol. 17, no 6, p.

1423.

WINSYS 2018 - International Conference on Wireless Networks and Mobile Systems

302