INSTAGRAM FAKE PROFILE DETECTION USING MACHINE LEARNING

Abstract

One of the major issues with Online Social Networks (OSNs) is fake interaction, which is used to artificially boost an account's popularity. The identification of fraudulent involvement is essential to prevent company losses, inaccurate audience targeting in advertising, inaccurate product prediction algorithms, and a negative impact on the atmosphere of social networks. This initiative focuses on the identification of automated and phoney accounts that cause phoney activity on Instagram. We are aware of no publicly accessible dataset for automated and phoney accounts. Two datasets have been created for this purpose in order to detect automated and phoney accounts. Machine learning approaches including Naive Bayes, logistic regression, support vector machines, and neural networks are used to find these accounts. Due to the dataset's unnatural bias, a cost-sensitive evolutionary algorithm is also used for the detection of automated accounts. Smote Nc technique is used to address the unevenness issue in the fictitious dataset. The results are 86% and 96%, respectively, for the automated and false account detection issues. Nowadays, the majority of people utilise social networking sites on a daily basis. Numerous people create profiles on social networking websites every day and connect with others there, regardless of their location or time. False identities are prevalent in different types of crime, as well as advanced persistent threats. Social networking site users not only benefit from them but also worry about the security of their personal information. We must first identify the user's social network accounts in order to determine who is disseminating threats in social networks. Based on the classification, it is required to distinguish between real and phoney profiles on social media. Several categorization techniques have traditionally been used to identify phoney social media accounts. However, there are ways to improve social media's ability to identify phoney profiles. The suggested effort uses technology and machine learning to boost the percentage of predicted phoney profiles.