Mobile Ad-hoc Networks working group Errong Pei Internet Draft School of Communication and Information Engineering Chongqing University of Postsand Telecommu. December 2017 Intended status: Informational Expires: June 2018 Anti-SSDF framework for Cognitive Sensor Networks draft-pei-antissdf-00.txt Status of this Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html This Internet-Draft will expire on June 15, 2009. Copyright Notice Copyright (c) 2017 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this Pei Expires June,2018 [Page 1] Internet-Draft Anti-SSDF framework December 2017 document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Abstract In cognitive sensor networks, the cooperative spectrum sensing can effectively improve the accuracy of spectrum sensing. However, it is also facing with the security issues caused by attacks of potential malicious users. Lots of algorithms are proposed in existing literatures to cope with the potential attacks. However, the traditional strategies are mainly suitable for relative simple and ideal attack such as the all "1" or all "0" attack. Therefore, we define a more universal anti Spectrum Sensing Data Falsification (SSDF) framework in the document, which can better cope with the probabilistic attacks and random attacks which are closer to the reality. Table of Contents 1. Introduction ................................................ 2 2. Conventions used in this document............................ 4 3. The data reconstruction based anti SSDF framework ............ 5 3.1. The detection framework of malicious users .............. 5 3.2. The processing framework of malicious users ............. 6 4. Formal Syntax ............................................... 7 5. Security Considerations...................................... 7 6. IANA Considerations ......................................... 7 7. Conclusions ................................................. 7 8. References .................................................. 8 8.1. Normative References.................................... 8 8.2. Informative References.................................. 8 1. Introduction Spectrum sensing is the basis and prerequisite of cognitive sensor networks. However, due to the shadow effect and multipath fading of the signal as well as the limitation of the calculation performance of single cognitive sensor node (or user), there exist big errors in the result of the spectrum sensing. The cooperative spectrum sensing among multiple cognitive sensor nodes is thus proposed to enhance the spectrum sensing performance of single cognitive sensor node. More specifically, multiple cognitive sensor nodes detect the spectrum simultaneously, one of which is responsible for fusing all detected data and making the final decision. As a key technology of cognitive radio, cooperative spectrum sensing not only has excellent sensing performance, but also can reduce the sensitivity Pei Expires June, 2018 [Page 2] Internet-Draft Anti-SSDF framework December 2017 requirements of detection hardware and further reduce deployment cost of the cognitive sensor networks. However, there are serious security issues in the cooperative spectrum sensing while the accuracy of spectrum sensing can be improved effectively. In the cooperative spectrum sensing, some malicious nodes (users) among the cognitive sensor nodes (users) may deliberately influence the spectrum sensing process through forgery, deception, flooding and gang cooperation, which makes the fusion center get the wrong sensing information. Based on the wrong sensing information, the fusion center may make wrong final decision, and further obtain wrong detection result. Thus the cognitive sensor nodes (users) cannot faithfully perform spectrum switching according to the external environment, and the channel allocation will be controlled and used by the attacker. This kind of attack method, which interferes with the normal operation of the data fusion center by sending the fake sensing data to the fusion center, is referred to as Spectrum Sensing Data Falsification (SSDF) attack. As a considerably important topic in cognitive sensor networks, the security issues have received increasing attention. Among them, the SSDF attack is a major security threat in cognitive radio networks. Currently, the anti SSDF algorithm is mainly divided into two steps: the first step is to detect malicious users, and the second step is to how to process the malicious data in the fusion center. In the detection of malicious users, some literatures proposed to detect malicious users by comparing the local judgment results with the final results in the fusion center. It is assumed in this kind of detection method that the malicious users are sending the wrong data. The implementation is thus relatively simple. But in reality, malicious users may only send the wrong data with a certain probability, thus the proposed algorithms cannot effectively cope with the SSDF attack. Other literatures proposed to detect malicious users through estimating the deviation value of the sensing data from their mean value. But in the scenarios with more malicious users, a more robust estimation method of the mean value is urgently needed. In the processing of malicious data, there are generally three processing methods of malicious data according to the existing literatures. The first method is to delete the malicious data directly. The method is the simplest. Obviously, the final judgment of the sensing result can be greatly affected due to the lack of the deleted malicious data. The second is to remove the malicious data, and then use the mean value of the entire dataset to replace it. Although the method is better than the first, the differences of Pei Expires June, 2018 [Page 3] Internet-Draft Anti-SSDF framework December 2017 different sensing data is erased as well, which can still affect the final judgment of sensing results. The third is to give different weight to different sensing results of multiple cognitive sensor nodes (users) according to certain an algorithm to neutralize malicious data. If the local judgment results of a cognitive user are consistent with the final results, more trust/weight is given, and vice versa. By employing the method, the attacks of malicious users can be effectively inhibited. However, in some special occasions, more complex attack modes might be used by malicious users to deceive the detection mechanism. For example, a malicious user deliberately operates as a normal user within some time, and then starts to attack the system after acquiring enough trust of the fusion center. The attackers can initiate an SSDF attack in a pre- designed attack mode with a quite high trust degree. Therefore, the attacks of malicious users are hard to be inhibited effectively by means of the user's entire historical performance or simply the accumulated reputation value. Furthermore, the weighted algorithm is not fair for the normal users because part of the sensing data of malicious users is still remained in the final fusion process. Additionally, the weighted malicious data is not necessarily the normal and effective data. In the case, the malicious user still seriously affects the detection performance of the sensing system. Besides, in the case of malicious data seriously deviating from the mean value, the effect of the weight of sensing data is more fragile which can greatly affect the final judgment. Considering the insufficiency of traditional anti-SSDF strategies, we define an anti-SSDF framework with more universality in the document: the data reconstruction based anti-SSDF framework. 2. Conventions used in this document "SSDF" indicates Spectrum Sensing Data Falsification "OGK" indicates Orthogonalized Gnanadesikan Kettenring "Cognitive users" also indicates the sensor nodes "Detect" also indicates sense The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119]. In this document, these words will appear with that interpretation only when in ALL CAPS. Lower case uses of these words are not to be interpreted as carrying significance described in RFC 2119. In this document, the characters ">>" preceding an indented line(s) indicates a statement using the key words listed above. This Pei Expires June, 2018 [Page 4] Internet-Draft Anti-SSDF framework December 2017 convention aids reviewers in quickly identifying or finding the portions of this RFC covered by these keywords. 3. The data reconstruction based anti SSDF framework 3.1. The detection framework of malicious users In order to find malicious data, the mean value mu and variance sigma^2 of the dataset need to be calculated. Then the malicious data can be distinguished through its deviation degree. At present, the estimation algorithm is shown as the following formula: mu^=1/n*(sum(mu_i)) sigma^2=1/n*(sum(mu_i-mu^)^2) This kind of estimation algorithm is simple and convenient. However, there are big errors in the estimation algorithm when there exist some malicious data, especially when malicious data is too much. Therefore, in order to detect malicious users more accurately in more universal cases, a kind of mean and variance value estimation algorithm with robustness is necessary. Based on this, the Orthogonalized Gnanadesikan Kettenring(OGK) algorithm is considered to be used in the document to estimate mu and sigma^2, which makes the estimation value more close to the actual value. OGK algorithm was firstly proposed by Maronna and Zamar in 2002. This kind of algorithm can provide a relative accurate estimated value when there exist part of malicious value. It has better robustness, stronger malicious data tolerance ability and lower algorithm complexity. The dataset of U={mu_1,mu_2,...,mu_n} is seen as a sample with single variable in OGK algorithm, and meets normal distribution with mu as mean value and sigma^2 as variance. Mu and sigma^2 are calculated according to the following formula: Mu^=sum(mu_i*W(v_i))/sum(W(v_i)),v_i=(mu_i-mu_0)/sigma_0 sigma^2=(sigma_0^2/n)*sum rho((mu_i-mu^)/sigma_0) Where the weighted function W(x) and rho(x) function can be described as, respectively, W(x)=(1-(x/c_1)^2)^2, (|x|<=c_1) Rho(x)=min(x^2,c_2^2) Pei Expires June, 2018 [Page 5] Internet-Draft Anti-SSDF framework December 2017 Where c_1 and c_2 are turning coefficient, which can be determined according to actual situations. Mu_0 and sigma_0 are median and absolute median deviation in dataset U, respectively. The dataset U={mu_1,mu_2,,mu_n} can be standardized into Y={y_1,y_2,...,y_n}, where y_1=|mu_1-mu^/sigma^|,...,y_n=|mu_n- mu^/sigma^|. Let V_i=y_i(i=1,...,n) denote the residual error. When the residual error of a data meets V_i>3*sigma (sigma is the standard error that can be obtained from the calculation based on the Bessel formulas), the data will be considered as malicious data (the corresponding user is a malicious user). 3.2. The processing framework of malicious users The k-means clustering algorithm in data mining is adopted in the document to process malicious data. In 1976, the K-means clustering algorithm was proposed by Macqueen (J.B.Macqueen) to process data clustering problem. This kind of clustering algorithm is widely applied because it's simple and convenient. It also has high scalability, good reliability and high efficiency. Therefore, it is often applied in science and industry areas. The processing flow of K-means clustering algorithm is shown as below: Input: dataset X={x_1,x_2,...,x_n}, cluster number k, Output: k class clusters C_j,j=1,2,...,k. [step1] Let I=1, randomly select k data points as initial cluster centers of k class clusters, m_j(I), j=1,2,...,k; [step2] Calculate the Euclidean distance of each data point to the k class clusters, d(x_i,m_j(I)), i=1,...,n ,j=1,...,k. If satisfied d(x_i,m_j(I))=squr((x_i-m_j(I))^2)=min{ d(x_i,m_j(I))}, then x_i_is part of C_J; [step3] Calculate k new cluster centers: m_j(I+1)=(1/N_j)*sum(x_i), x_i is part of C_j, j=1,...,k; [step4] Judge: if m_j(I+1)is not equal to m_j(I), j=1,...,k, then I=I+1, return to step2, otherwise, the calculation finishes. Pei Expires June, 2018 [Page 6] Internet-Draft Anti-SSDF framework December 2017 After the clustering grouping of normal data is completed, the data is divided into k clusters and each cluster has a cluster center m_j(I), j=1,2,...,k. According to the relativity of cognitive user data with geographic position, the malicious users can be categorized into their nearest cluster (denoted by C_jM) to their self, and then the malicious data can be replaced with the cluster center value m_jM(I) of the cluster C_jM. After completing the above data replacement, the final fusion value can be calculated as follows: T=(mu_1+,...,+mu_(n-M))+(m_1(I)+...+m_M(I)) The final judgment of the cooperative spectrum sensing can be made as follows If T is more than eta, then the detected channel is occupied; If T is less than eta, then the detected channel is idle. Where T represents the final fusion value, n represents the number of cognitive users, M represents the number of malicious users among the cognitive users, mu_i represents sensing data sent to the fusion center by normal cognitive users, m_h(I),h=1,...,M represents the latest sensing data of malicious users after data reconstruction(ie. cluster center value of the nearest cluster to their self), eta represents the judgment threshold of the fusion center. 4. Formal Syntax The following syntax specification uses the augmented Backus-Naur Form (BNF) as described in RFC-2234 [RFC2234]. 5. Security Considerations The document introduces a new anti SSDF framework for the cognitive sensor networks. The framework does not inherently introduce any additional threats. 6. IANA Considerations This document has no actions for IANA. 7. Conclusions The document defines a more universal anti DDSF framework. Pei Expires June, 2018 [Page 7] Internet-Draft Anti-SSDF framework December 2017 8. References 8.1. Normative References P.Kaligineedi, M. Khabbazian, and V. Bhargava, ''SecureCooperative Sensing Techniques for Cognitive Radio Systems,''in Proc. IEEE ICC, 2008, pp. 3406-3410. F. Penna, Y. Sun, L. Dolecek, and D. Cabric, "Detecting and counteracting statistical attacks in cooperative spectrum sensing," IEEE Trans.Signal Process., vol. 60, no. 4, pp. 1806-1822, Apr. 2012. 8.2. Informative References VI T. Zhang, R. Safavi-Naini, and Z. Li, "ReDiSen: Reputation-based securecooperative sensing in distributed cognitive radio networks" inProc. IEEE ICC, Budapest, Hungary, Jun. 9-13, 2013, pp. 2601-2605. This document was prepared using 2-Word-v2.0.template.dot. Authors' Addresses Errong Pei School of Communication and Information Engineering Chongqing University of Posts and Telecommunications Nanan Dist., Chongqing, China
Phone: 008613638323589 Email: peier@cqupt.edu.cn Pei Expires June, 2018 [Page 8]