International Journal of Engineering Trends and Technology (IJETT) – Volume17 Number6–Nov2014 A Manageable Framework for Preserving Exclusive Data in Social Networks 1 A.v.swathi,2C.P.V.N.J Mohan Rao 1 Final M.Tech Student, 2Professor 1,2 Dept of Computer Science and Engineering, Avanthi Institute of Engineering & Technology - Narsipatnam. Abstract: In social networks there are so many threats such as impersonation, data leakage. Most of the social networks are more worried about private data leakage. Some of them as intruders by deploying the security attacks like denial of service or hack the information. So we introduced a framework which contains agents on every Peer and they monitor every Peer to maintain confidentiality, anonymous etc. The main goals of these methods are to maintain privacy of private data. So we additionally introduced a classifier to predict that connected Peer is secure or then proceed to share data. I.INTRODUCTION Internet privacy refers as the right of self-privacy regards to the storing and provision to other parties and showing of information to oneself through the Internet. Web privacy is a part of system privacy and the privacy affects have been communicated from the starting of large scale computer sharing. Privacy can necessitate either Personally Finding Information (P-II) or non-PII information consists of site visitor behavior on a website. P-II is any information that can be used to find an individual. Consider an example age and address only could find an individual is without directly imparting their name and there are two factors are isolated enough to typically find a particular person. People with only a general affect for web privacy need not attain total anonymous. The web users may or may not protect their personal information using controlled disclosure. The revelation of IP addresses and the nonpersonally-identifiable profiling and it is same information would become acceptable trade-offs for the satisfaction and users could otherwise lose using the workarounds need to compress such details rigorously. On the other hand and some of the people desire much stronger privacy. In that situation they may try to attain Internet anonymous to ensure privacy and use of the Internet without giving any information to other members the ability to link the web bustle to personally able to find information of the web user. To keep their information private the people need to be very careful with the information what they submit. Filling of forms and buying in merchant sites that becomes trace and because the data was not private and present days the companies are used to send spam and advertising on same products. ISSN: 2231-5381 There are also so many government companies that protect user privacy and anonymous issues on the Internet. There is an article by the FTC, a number of indicators were brought to alert that helps user in internet to avoid possible personal information theft and other cyber-attacks. The process of blocking the usage of online is to wary and respected emails consists spam messages of financial details and those are creating and manipulating strong passwords. Sharing things on the Internet can be danger of attacks. There are some information shared on web is permanent and depending on service and privacy methods of some services offered online. It consists of comments on blogs and pictures. It is merged into cyberspace and once it is shared to anyone can find it and access it. Some employers started research a possible employee by searching online for the details of their online hiking patterns as possibly affecting the output of the success of the candidate. Companies are showing interest to watch what web sites people visit and then use that information for sample by sending advertising based on individual surfing history. There are several ways in which people can broadcast their information for once by use of channel and by sending the bank and credit card information to different websites. And directly remark the behavior such as browsing logs and search queries of the social website profile can be automatically processed to refer more unwanted details about an individual such as political and religious.[8] Those affected about web privacy frequently reproduce a number of privacy risks and events that can accommodate the privacy which may be confront through web use.[9] These ranges are from the collection of statistics on users to more viral acts such as the distributes spyware and deploying of various forms of software faults. There are several social websites try to protect the self-data of their payers. Consider an example the security settings are available to all registered users and they can block particular users from viewing their profile data they can choose their friends and they can terminate who has access to pictures and videos. The security settings are also available on other social networking sites such as Google Plus and Twitter as well as Facebook. The user can apply http://www.ijettjournal.org Page 262 International Journal of Engineering Trends and Technology (IJETT) – Volume17 Number6–Nov2014 such settings at the time of providing the personal information on the internet. Children also using the web including social networks in which risk their privacy and a cause for increasing affect among parents. They must be informed that about all these risks. Consider an example on Twitter threats include shortlisted links that lead them to harmful places. In their e-mail inbox and there are some threats that includes email scams and attachments that get them to install malware software and reveal their personal information. When using a smartphone and the threats include geo-location meaning that phone can detect where they are located and share that information in online for all to view. All users can protect themselves by updating the virus security which is using security settings and downloading the patches installing a firewall and controlling cookie programs using encryption method refuse the browser hijackers. removing the user details of the Peers before hosting the actual graph does not always provides privacy. Another one is to observe that the structural similarity of Peers neighborhood in the graph explains the the extent to which an individual in the network can be distinguished. This structural data is closely linked to the degrees of the Peers and their neighbors. The researchers propose an anonymous model for social networks and a graph achieves k-anonymous and for every structure query over the graph and there exists at least k Peers that match the query. The structure queries checks the presence of neighbors of a Peer or the structure of the sub graph in the territory of a Peer. The researchers mostly focus on giving a set of anonymous symptoms and studying properties of them and not on designing methods that bond the construction of a graph that obeys their anonymous requirements. III.PROPOSED WORK II. RELATED WORK Issues of social networking sites The appearance of the Web has caused social profile and is a growing affect for web privacy. Web is the system that provides the participatory information sharing on the web and in the social networking websites like Facebook and orkut. These social networking sites have seen a more famous in their popularity started from many years ago. These websites leads many people are giving their personal information on the internet. It has been a concept of discussion is held responsible for the collection and distribution of personal data. Some of them will say that it is the mistake of the social websites because they are storing large amounts of information. It is the users who are responsible for the situation because of the users themselves that provide the data in the first place. This can relate to the ever present the issue of how people regard social media websites. These are increasing the number of people that are finding the risks of keeping their self-information online and believe a website to keep it private. The information is not completely private for long time. There is an increase of risk because of people of below age of 20 are having easy internet access than ever before and so they place themselves in a position where all very easy for them to upload information. This is because of the fast involving of digital media and the people’s explanation of privacy is emerging very well and it is crucial to consider that when communicating in online. Sometimes it may be compulsory to take extra safeguard the situations where someone else may have a closer view on privacy rules. In recent work point out that the simple method of anonymous graphs by ISSN: 2231-5381 Our work is completely on attribute based method which has been used to point out security problems regarding to attacks in social networks. This paper implements novel method such as data mining and different agents to provide solutions for social networks. Our Proposed work privileges the three techniques to provide serve security solution to Self Peer, Friendly Peer and Central Networks. The below figure shows the architecture of the system to reduce the attacks in social networks. The following explains that each agent’s work in detail. A. Self-agent Self-agent is exists in each system and it collects data about its own from the application layer to routing layer. Our process provides solution in three techniques. 1. It always verifies own system and its habitat dynamically. And it uses classifier to find out the local anomalies. 2. Whenever the Peer wants to share the information from the Peer F to B. It broadcasts the message to E and A. Before it sends the message and it collects the information of the Friendly Peers (E &B) using mobile agent. Then it calls that the classification rules to find out the malicious attacks with help of training data. 3. It gives similar type of solution in entire the global networks. It has been presented completely in the following section. . 1) Self Peer - Self Agent is exists in the system and it verifies its own system frequently. If any attacker sends any data packet to collect information through this system and it calls the classifier process to find out the malicious http://www.ijettjournal.org Page 263 International Journal of Engineering Trends and Technology (IJETT) – Volume17 Number6–Nov2014 attacks. If an attack has been made then it will filter the attacker system such as the relation to that peer or node from the global networks. 2) Friendly Peer - Any system in the network shares any information to some other system and it broadcast through neighbor system. Before it transfers the message it sends mobile agent to the Friendly Peer and collects all information and it return back to the system and then calls classifier rule to find out the malicious attacks. If there is no malicious activity then it will forward the message to Friendly Peer. 3) Data Requirements - The requirements module is included for each anonymous user detection sub-system to collect the values of features for corresponding layer in any system. General profile is created using the data collected from the agents during the normal format. Attack data is collected during the attack process. 4) Preprocess – The verified data is collected in a file and it can be used for anonymous user detection. Preprocessing is a method to process the information with the test training data. In the complete layer anomaly detection systems are above mentioned pre-processing method is used. B. Cross feature evaluation for classifier sub model construction 1) Each character vector as ‘f’ in the training data set, compute classifier as C, for each feature fi using F={f1, f2... fi -I, fi+.-f k} - Ci is learned from the training data set using Bayes classification algorithm. The probability P. (fl, f2 ..., fi- i, f+1, ...,fk) is learned. 2) Calculate the average probability for every feature vector ‘f’, and it save in a probability distribution matrix as ‘M’. A decision threshold is ‘0’ is learned from the training data set and general profile is created using the threshold value. If the probability value is greater than threshold value it is conclude as normal, otherwise it is conclude as abnormal. Anomaly detection Algorithm:Input: Preprocessed training data, preprocessed testing data Output: Probability of anomaly 1) Read pre-processed data set file 2) Call Bayes classifier program for training the classifier for anonymous detection 3) Read the testing data file 4) Test the classifier process with the testing data file 5) Print the classification matrix to view the actual class vs predicted class 6) Percentage of anomaly is computed as follows Percentage = (Number of predicted abnormal class X 100)/Total number of traces Naïve Bayesian Classification Algorithm Algorithm to classify malicious agent Sample space: set of agent H= Hypothesis that X is a node P (H/Xi) is our confidence that Xi is an incoming node P(H) is Prior Probability of H and it is probability that any given training sample is an agent regardless of its anomaly or not anomaly behavior P(H/X) is a conditional probability and P(H) is independent of X Estimating probabilities P(H), P(Xi) and P(Xi/H) may be estimated from given training and testing data samples P(H|Xi)=P(Xi|H)*P(H)/P(Xi) Steps Involved: Each training data sample is of attribute type X= (xj) j =1(1….n), where xj is the values of X for attribute Aj 2. Suppose there are m decision classes Cj, j=1(1…m). P(Ci|X) > P(Cj|X) for 1<= j <= m, j>i i.e. classifier assigns X to decision class Cj having highest posterior probability conditioned on testing sample X The decision class for which P(Cj|X) is maximum is known as maximum posterior hypothesis of the sample. From Bayes Theorem 3. P(Xi) is constant and Only need be maximized. if class initial probabilities not known prior then we can assume all decision classes to be more equally likely decision classes Otherwise maximize the samples P(Ci) = Si/S 4. Naïve assumption for attribute independence P(X|Cj) = P(x1,…..,xm|C) = PP(xk|C) 5. To classify an unknown testing sample Xi, compute each decision class Ci and Sample X is assigned to the class iff ( Prob(Xi|Ci)P(Ci)> P(Xi|Cj) P(Cj) ). IV.EXPERIMENTAL ANALYSIS Source name ip /node Destination /node name ip Port no Type protocol of Number of packets (in bytes) Status 192.168.1.10 192.168.1.20 8081 TCP 56 Malicious 192.168.1.12 192.168.1.21 8082 TCP/IP 120 Not Malicious ISSN: 2231-5381 http://www.ijettjournal.org Page 264 International Journal of Engineering Trends and Technology (IJETT) – Volume17 Number6–Nov2014 192.168.1.11 192.168.1.20 8081 smtp 35 Malicious 192.168.1.19 192.168.1.28 8083 http 56 Malicious 192.168.1.16 192.168.1.25 8084 TCP 56 Not Malicious Our experimental results shows more efficient results than the traditional approach in finding anonymity problem and maintain authentication while communicate with other nodes, for experimental implementation we implemented through network programming in java. Initially source node connects to destination node through its ip address and port number and retrieves its meta data and forwards these sample to training dataset for analyzing the behavior by computing posterior probability [8] D. Hong, J. Sung, S. Hong, J. Lim, S. Lee, B. Koo, C. Lee, D. Chang, J. Lee, K. Jeong, H. Kim, J. Kim, and S. Chee, “HIGHT: A New Block Cipher Suitable for LowResource Device,” Proc. Eighth Int’l Workshop Cryptographic Hardware and Embedded Systems (CHES ’06), pp. 46-59, 2006. [9]Pogue, David (January 2011). "Don't Worry about Who's watching". Scientific American 304 (1): 32. doi:10.1038/scientificamerican0111-32. REFERENCES [10]"The [1] Internet Traffic Classification by Aggregating Correlated Naive Bayes Predictions. By Jun Zhang, Member, IEEE, Chao Chen, Yang Xiang, [2], “Internet traffic classification using bayesian analysis techniques,” by A. W. Moore and D. Zuev [3] SPINS: Security Protocols for Sensor Networks ADRIAN PERRIG, ROBERT SZEWCZYK, J.D. TYGAR, VICTOR WEN and DAVID E. CULLER [4] M. Hay, G. Miklau, D. Jensen, P. Weis, and S. Srivastava.Anonymizing social networks. Technical Report 07-19, University of Massachusetts Amherst, 2007. [5] J. He, W. Chu, and V. Liu.Inferring privacy information from social networks.In Mehrotra, editor, Proceedings of Intelligence and Security Informatics, volume LNCS 3975, 2006. [6] T. Joachims. Training linear SVMs in linear time. In ACM SIGKDD In- ternational Conference On Knowledge Discovery and Data Mining (KDD), pages 217 – 226, 2006. Value of Privacy by Bruce Schneier". Schneier.com. Retrieved 2011-11-25. BIOGRAPHIES Dr. C.P.V.N.J Mohan Raois Professor in the Department of Computer Science and Engineering, Avanthi Institute of Engineering & Technology - Narsipatnam. He did his PhD from Andhra University and his research interests include Image Processing, Networks, Information security, Data Mining and Software Engineering. He has guided more than 50 M.Tech Projects and currently guiding four research scholars for Ph.D. He received many honors and he has been the member for many expert committees, member of many professional bodies and Resource person for various organizations. A.v.swathi completed B.techc.s.e from JNTU(2002-2006) from Al - Ameer college for engineering & IT Worked for Option I Technologies,Hyderabad as SAP Trainer. (2006-2008). Worked as Assistant professor for Gayatri college of engineering(C.S.E) Visakhapatnam(2008-2010)M.techc.s.e JNTUK (2012 2014) from Avanthi college of engineering –Narsipatnam. [7] A. Perrig, R. Szewczyk, J. Tygar, V. Wen, and D. Culler, “SPINS: Security Protocols for Sensor Networks,” Wireless Networks, vol. 8, no. 5, pp. 521-534, 2002. ISSN: 2231-5381 http://www.ijettjournal.org Page 265