A METHOD FOR IDENTIFICATION AND DETECTION OF FAKE PROFILES ON ONLINE SOCIAL NETWORKS

Abstract

A METHOD FOR IDENTIFICATION AND DETECTION OF FAKE PROFILES ON ONLINE SOCIAL NETWORKS The present disclosure provides a method for identification and detection of fake profiles on online social networks. The present invention is solving the problem of fake accounts and profiles created by humans or bots on online social networks like Facebook, Instagram, Tweeter etc, such fake accounts are used to disseminate fake rumours and engage in illicit activities like theft and phishing and create a potential threat to multiple community and sometime mentally disturb users. The present invention employs a combination of machine learning and deep learning modules that together help in accurately identifying and detecting fake users’ profiles and preventing such fake profiles from spreading misleading information. The method involves collecting dataset from the online social platform, subjecting it for pre-processing followed by extracting informative features from pre-processed data set for distinguishing between fake and genuine accounts then processing the eextracted data set by employing Natural Language Processing (NLP), classifying pre-processed dataset by splitting it into training and testing data, classifying, the dataset to detect fake accounts by applying Recurrent Unit GRU and Long Short-Term Memory LSTM and combining predictions from multiple classification algorithms to create a hybrid classification system i.e. a hybrid classification algorithm (GUR + LSTM).

Specification

We Claim:
1. A method for identification and detection of fake profiles on online social
networks, comprising the steps of:
i. collecting, a plurality of data set from an online community platform,
the said data set consist of genuine and fraudulent users' profiles;
ii. pre-processing, the data set by applying a plurality of computer
generated instruction with the help of a computing device to identify and
remove missing values in the users' profiles, ensuring consistency and
preventing redundancy for subsequent steps;
iii. identifying and extracting, informative features from pre-processed data
set for distinguishing between fake and genuine accounts by performing
feature selection techniques, said extracted features include: account
creation date, tweet frequency, follower-to-following ratio, engagement
metrics (e.g., likes, retweets), and linguistic patterns;
iv. processing, the identified and extracted features data set by employing
Natural Language Processing (NLP) so as to process the textual content
of tweets or reels, the NLP performs removal of stop words, and
lemmatization followed by applying text normalization, and extracting
features from the processed text, such as bag-of-words representations,
TF-IDF values, or word embeddings;
v. classifying, pre-processed dataset by splitting it into training and testing
data using machine learning classifiers, the machine learning classifiers
such as Logistic Regression (LR), Random Forest (RF), and Support
Vector Machines (SVM) are applied to the training dataset;
vi. classifying, the dataset to detect fake accounts by applying Recurrent
Unit GRU and Long Short-Term Memory LSTM, and training a deep
learning module based on pre-processed dataset and evaluating its
performance on the testing set;
vii. combining, predictions from multiple classification algorithms such as
machine learning and deep learning models, to create a hybrid
classification system i.e. a hybrid classification algo

Documents