Description

Abstract - The social network, a crucial part of our life is plagued by online impersonation and fake accounts. This project is to determine whether the profile is fake or normal account. Now a days social media is playing crucial part in every one life undergoing with lot of trouble due to fake accounts. In order to overcome this we propose a model that could be used to classify an account as fake or genuine. By using support vector machine as a classification technique which can process large dataset of accounts at once, eliminating the need to evaluate each account manually. As this is an automatic detection method, it can be directly applied to social networks  which consists of millions of profiles.

Keywords – Support vector machine, Classification,  fake user detection,   social media.

l . INTRODUCTION

      In the present age, the public activity of everybody has become related with the online interpersonal organizations. Including new companions and staying in touch with them and their updates has gotten simpler. The online social communities have much influence on the science, instruction, grassroots arranging, work, business, and so forth. Scientists have been concentrating these online informal organizations to see the effect they make on the individuals.

Educators can arrive at the understudies effectively through this online organization, instructors these days are getting themselves recognizable to these destinations bringing on the web study hall pages, giving schoolwork, making conversations, and so on which improves training a great deal. The businesses can utilize these network   communication locales to utilize the individuals who are capable and keen on the work, their record verification should be possible effectively.

In this paper by using classification algorithm which can determine the fake account in social network .support vector machine is one of the classification technique which is considered as accurate algorithm among classification algorithm.

ll. EXISTING SYSTEM

•        The existing Naive Bayes algorithmic program  has less accuracy.

•        Particularly since late 2016 throughout the Presidential election, the question of determinant 'fake news' has additionally been the topic of specific attention among the literature. 

•        Conroy, Rubin  outlines many approaches that appear promising towards the aim of absolutely classify the dishonest articles.

•        They state  that easy content-related to n-grams and shallow parts-of-speech (POS) tagging have verified deficient  classification task, usually failing to account for necessary context data.

•        Rather, these strategies are shown helpful only in cyclic process with additional advanced strategies of study.

lll. PROPOSED SYSTEM

•        Classification starts from the selection of profile that must to be classified.

•         Once the profile is chosen , the needed features are extracted for the purpose of classification.

•         The extracted features are then fed to trained classifier.

•         Classifier is trained frequently as new information is fed into the classifier.

•         Classifier then determines whether the profile is genuine or fake by using support vector machine classification (svm) algorithm

•        The results of  classification algorithmic rule is then verified and feedback is fed into the classifier.

•        As the range of training information will increase the classifier becomes additional and a lot of correct in predicting the pretend profiles.

lV. LITERATURE SURVEY

 In 2017 Dr Vijay Tiwari proposed Analysis and detection of fake profile over social network in this paper Analysis of user metadata and machine learning techniques has an edge over the graph technique (logistic regression) in detecting the fake profile.[1]

In 2018 Estée van der Walt,  jan eloff proposed Using Machine Learning to Detect Fake  Identities: Bots vs Humans in this paper Filtering of fake accounts carried out through supervised and unsupervised algorithms.[2]

In 2018 Akshay Jain , Amey Kasbe  proposed Fake News Detection Naive Bayes classification model to predict whether a post on social account will be labeled as REAL or FAKE. Web  scrapping method is used to manage large data sets [3]

In 2015 Haoran Xu and Yuqing Sun proposed Identify User Variants Based on User Behavior on Social Media It Studies  the characteristics of user behaviors on social media and introduce two concepts visibility And distingushibility to preliminarily quantify whether a fake user can be identified  [4]

In 2018 Shivangi Gheewala , Rakesh Patel proposed Machine learning based twitter spam account detection: a review Machine learning techniques categorized spam detection into syntax analysis and feature analysis which uses statistical features for spam detection [5]

In 2019 Ahmad Almogren , faiza masood proposed Spammer Detection and Fake User Identification on Social Networks Naïve Bayes, random forest, bayes betwork, K-nearest neighbor, clustering, and decision tree algorithms are used for predicting and analyzing spams on Twitter with different classes of categorization  [6]