Design and modeling an Adaptive Neuro-Fuzzy Inference System (ANFIS) for the prediction of a security index in VANET

https://doi.org/10.1016/j.jocs.2020.101234Get rights and content

Highlights

  • This work presents one of the first proposal of soft computing for the prediction of attacks in this type of network. Fuzzy inference systems provide an intuitive and high-level mechanism for representing knowledge.

  • Our model shows how by the use of neural networks and fuzzy logic, we can obtain a security index as a measure of protection and possible attack prediction. One of the most important avantage is that the ANFIS scheme is computationally efficient.

  • This system increases the dynamic performance and provides good stabilization when there is a sudden fluctuation in one of the system parameters. In our specific case, it was observed how the security index obtained responds according to the intersection of the variables “Transmission” and “Lost Packages”.

  • Another advantage of the model proposed is the learning capability using the neural networks. ANFIS system handles more complex parameters that was obtained from a variables’ reduction with the APC. This proves the effectiveness of the sudden variation of our security index from a normal behavior of the network and its effects when a vehicle is facing an attack.

  • When integrated to an anomaly based intrusion detection system, this predictive model makes it possible to improve detection patterns of attacks in real time.

Abstract

Vehicular Ad hoc NETworks (VANET) allow communications between vehicles using their own connection infrastructure. There are several advantages and applications in using this technology and one of most significant is road safety. As in most other networks, it is not only important to guarantee the transport but also the security of information. Security in VANET is a big challenge because there are different types of attacks that endanger communications of moving vehicles. This paper proposes an applied Adaptive Neuro-Fuzzy Inference System (ANFIS) to obtain a prediction model of security index in VANET. The research process starts with network simulations to obtain a database of occurrences of attacks. Then, this latter is prepared and analyzed statistically. Finally, using MATLAB toolbox, we show the proposed model of security level that allows estimating the network vulnerability in the event of an attack.

Introduction

The era of communication advances faster, computers, telephones and millions of devices are connected to transmit and receive information. VANET is a kind of mobile network (MANETs) that allows communication with or without infrastructure installed at the edges of the roads [1]. In this type of network, the nodes are vehicles and one of the main features is that its topology can change quickly. The main contribution in this type of network is the development of applications in the framework of Intelligent Transport Systems (ITS) [2], which are used to provide and improve traffic information and in this way, it increases road safety and transport efficiency.

The network security is very vulnerable, there are several technical challenges in VANET, the high mobility, routing, dynamic topology, loss of information through its wireless connection, among others [3]. Main security problems can occur during the transmission of information. Different types of attacks can violate and/or interrupt the connection and there are flaws or anomalies typical of the communication system. Security protocols, formalization of standards and different analysis of attacks, have been proposed to improve VANET security, but the field is still large to explore [4].

Following this line of research, this work is based in soft computing techniques and artificial intelligence. We select the fusion of fuzzy logic and neural networks as a method to design a model that predicts the security level against attack possibilities in VANET.

It is very important to highlight that to date; this is the first work that implements fuzzy logic techniques as a predictive method for the detection of attacks or anomalies in this type of network. To achieve a good model for this level of security, statistical and data processing tools were needed too.

As in all emerging technology, uncertainty and the impression of these types of system is a variable that significantly influences the progress of its development. The adaptive neuro-fuzzy inference system (ANFIS) is characterized by incorporating aspects of neural networks and fuzzy logic. The first one takes advantage of the ability to learn, as well as the ability to generalize. From the fuzzy logic is obtained the logical reasoning based on rules of inference, thus contributing, a powerful tool that allows to operate with linguistic variables and incorporates a broader treatment. The construction of models based on these two areas has proven to be an efficient mechanism when modeling real systems.

The different integrators used in ANFIS models with neural network algorithms focus on their application to prediction. Thus, the objective is to minimize prediction errors. On the various works carried out to which we have access, shows us a good prediction performance with a fairly low error rate and a better accuracy. This motivated us to use an ANFIS model in the detection of intrusions into VANET networks.

The remainder of this paper is organized as follows. In Section 2, a Literature review (state of art) about security issues and attacks in VANET is presented. Section 3 presents the methodology and tools used in this work. In Section 4, we present the analysis process and data treatments. Section 5 is focused on our proposal for a model aiming at measuring a security index in VANET. Finally, a discussion about results is presented in Section 6.

Section snippets

State of art

The use of different wireless technologies in vehicular environments has been studied since the 1970s [5]. As a complex system, it is very important to understand each component of VANET. Several works have been presented and served as reference for the definition of the main concepts in vehicular networks operations. In [1], that can be established by short and medium range communication based on WLAN (Wireless Local Area Network) technology. For VANET, IEEE 802.11p and IEEE 1606.4 standards

Methods and tools

This work aims to define a predictive model based on adaptive neuro-fuzzy systems (ANFIS) in order to obtain a security index that will be used in intrusion detection systems for vehicular networks. To understand our approach, we present in this section a summary of different models we used. The ANFIS is a hybrid artificial intelligence methodology that combines the fuzzy logic with the ability of neural networks to detect patterns in data and to learn from the relationships in different

Analysis and data treatment

We present in this section the details of the different steps required to obtain the data and the preparation of the variables to be included in our predictive model.

ANFIS model for the prediction of a security index in VANET

In this section, we will define and build the ANFIS model using MATLAB Toolbox (Fig. 10). We start with the ANFIS training that is performed to obtain the optimized values of all its modifiable parameters of a diffuse inference system (FIS). The basic flow diagram of computations in ANFIS is shown in Fig. 9.

The architecture model is presented in Fig. 10. We have two input variables, which are the same as those obtained in the previous section. Then, we have an inference engine of Takagi–Sugeno

Results discussion

Our aim was to use a predictive model to obtain a security index in VANET from simulation results. The input of the model is the message transmission and the lost packets during the communication. So, the output is the security index that allows to detect if the network is attacked or not. To test the predictive capability of the model, we developed it first; this later is then ‘trained’ on one set of data and it is ‘tested’ on previously unseen data we collected independently. Finally, the

Conclusion

Vehicular Ad hoc NETworks present great challenges. To ensure their implementation and effectiveness, many studies and research have been carried out these last years. This work presents one of the first proposal of soft computing for the prediction of attacks in this type of network. Fuzzy inference systems provide an intuitive and high-level mechanism for representing knowledge. Our predictive model shows how by the use of neural networks and fuzzy logic, we can obtain a security index as a

Conflict of interest

This work was done for the graduation of university master. My two students, Mrs. Pereira and Mr. Lahrouni and I, as supervising teacher of this project, did not benefit financially from this project. This project was completed with the assistance of a NSERC federal grant.

Funding

Natural Sciences and Engineering Research Council of Canada, Grant/Award Number: RGPIN 23972-2013.

Prof. Boucif Amar Bensaber received the PhD degree in computer science from René Descartes University, Paris V, France, in 1998 and the M.S. in Computer Science, from René Descartes University in 1993. He works as research scientist at Research Center for Effective Diagnostics, (CRED), Sherbrook, Canada from 1999 to 2000. He is currently Full Professor in Computer Science at University of Quebec at Trois-Rivières, Quebec, Canada. He has been head of the computer science section from 2005 to

References (41)

  • L. Le, A. Festag, R. Baldessari, W. Zhang, Vehicular wireless short-range communication for improving intersection...
  • I.S. Association, et al., Draft standard for wireless access in vehicular environments (wave)-multi-channel operation,...
  • S.N. Pathak et al.

    Secured communication in real time VANET

  • J.T. Isaac et al.

    Security attacks and solutions for vehicular ad hoc networks

    IET Commun.

    (2010)
  • J.M.d. Fuentes, A.I. González-Tablas, A. Ribagorda, Overview of Security Issues in Vehicular Ad-hoc...
  • H. La Vinh et al.

    Security attacks and solutions in vehicular ad hoc networks: a survey

    Int. J. AdHoc Netw. Syst. (IJANS)

    (2014)
  • R.S. Raw et al.

    Security challenges, issues and their solutions for VANET

    Int. J. Netw. Secur. Appl.

    (2013)
  • A. Aijaz et al.

    Attacks on inter vehicle communication systems—an analysis

    Proc. WIT

    (2006)
  • J. Grover, M.S. Gaur, V. Laxmi, Multivariate verification for Sybil attack detection in VANET, Open Comput. Sci....
  • M.N. Mejri et al.

    Gdvan: a new greedy behavior attack detection algorithm for VANETs

    IEEE Trans. Mob. Comput.

    (2017)
  • Cited by (26)

    • IASMFT: intelligent agent simulation model for future trading

      2024, International Journal of Information Technology (Singapore)
    • Multi-Agent System for Portfolio Profit Optimization for Future Stock Trading

      2024, Karbala International Journal of Modern Science
    View all citing articles on Scopus

    Prof. Boucif Amar Bensaber received the PhD degree in computer science from René Descartes University, Paris V, France, in 1998 and the M.S. in Computer Science, from René Descartes University in 1993. He works as research scientist at Research Center for Effective Diagnostics, (CRED), Sherbrook, Canada from 1999 to 2000. He is currently Full Professor in Computer Science at University of Quebec at Trois-Rivières, Quebec, Canada. He has been head of the computer science section from 2005 to 2009 and from 2019 to 2021. His research interests include Security, Networking, Sensor Networks, Communication Protocols, Wireless and Mobile Networks, Vehicular Ad Hoc Networks, and Medical informatics. He has multiple publications in international sources. He has supervised 3 PhD students and 50 Master students. He has been guest editor of a volume in Journal of Computational Science-Elsevier in 2016. He is a reviewer for many international Journals (International Journal of Communication Systems, Wiley; Vehicular Communications Journal, Elsevier; Security and Privacy Journal, Wiley; IEEE Communications Magazine; ACM Computing Surveys Journal; IEEE Transactions on Computers Journal, etc.). He is also a TPC member for many conferences (IEEE ICC, Globecom, IEEE/WCNC, IWCMC, etc.). He was the Co-chair of 2020 IEEE Global Communications Conference: Communications Software, Services and Multimedia Apps.

    Caroly Pereira received her master's degree in mathematics and computer science in 2019 with the mention Excellent. She worked in the development of solutions for the detection and prevention of attacks on Ad hoc vehicle networks (VANETs). From 2012 to 2017, she was a counselor and then director of an insurance company in Venezuela. Since 2017, she has worked as a mathematician with the company Bluberi Canada in the development of game software.

    Youssef Lahrouni received his master's degree in mathematics and computer science in 2017. He worked on Mathematical Methods against Denial of Service (DoS) Attacks in VANET. Since 2017, he is working at CGI Canada programmer analyst.

    View full text