Artificial Neural Networks for Social Media Fake Account Detection

Abstract

Artificial Neural Networks for Social Media Fake Account Detection ABSTRACT There are significant issues with platform integrity, user privacy, and trust when fraudulent accounts proliferate on social networking platforms. These phoney accounts are commonly used to influence public opinion, commit financial fraud, disseminate misleading information, and take part in phishing assaults. The increasingly sophisticated tactics employed by those who create phoney accounts have proven too tough for traditional detection methods, such as rule-based algorithms, to keep up with. This invention uses Artificial Neural Networks (ANNs) to provide a sophisticated approach for detecting fraudulent accounts by looking at user behaviour, interaction patterns, account data, and content properties. The ANN-based system integrates many neural network designs, including Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), and Recurrent Neural Networks (RNNs), to handle a range of data inputs, including profile metadata, images, and sequential user behaviours. Because it minimises false positives while attaining high detection accuracy through adaptive learning and real-time monitoring, the approach is scalable for large social media networks. The proposed solution strengthens platform security and user authenticity while also increasing detection efficacy. Our AI-driven approach ensures safer online interactions and greater digital trust by providing a robust, automated, and adaptable solution to the persistent problem of fraudulent accounts.

Specification

Description:FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENT RULES, 2003
Complete Specification
(See section 10 and rule 13)
Title of the Invention: Artificial Neural Networks for Social Media Fake Account Detection
2. Applicants

Name
Nationality
Address
Dr.R.S.Arunkumar Indian Associate Professor, Department of Computer Science and Engineering, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Science (SIMATS), Tamilnadu-602105, India.
Miss. Shradha Dubey Indian Assistant Professor, Department of Computer Science, Amity University, Gwalior, Madhya Pradesh - 474001, India.
Dr.J.Jeyabharathi Indian Associate Professor, Department of CSE, Kalasalingam Academy of Research and Education, Krishnankoil, Virudhunagar, Tamilnadu - 626126, India
M. Rajeshwari Indian Assistant Professor, Department of Management Studies, PSR Engineering College, Sivakasi, Tamilnadu- 626125, India.
Dr Ritika Indian Assistant Professor, Department of CSE, Apeejay Stya University, Palwal - Sohna Rd, Gurugram, Haryana-122103, India.
Krishna B Koppa Indian Associate Professor, Department of Marketing, Faculty of Management Studies, CMS Business School, JAIN (Deemed-to-be University), Sheshadri, Road, Bangalore, Karnataka- 560098, India.
Manjunath B T Indian Assistant Professor, Department of Mathematics, BMS College of Engineering, Amrita School of Engineering, Amrita Vishwa, idyapeetham, Karnataka-560011, India.
Dr. Leena Chhabra Indian Assistant Professor, Department of Law, GD Goenka University, GD Goenka Educational City, Sohna Gurgaon Road, Sohna, Haryana-122009, India.
A. Vani Lavanya Indian Assistant Professor, Department of Computer Science and Engineering, St. Joseph's Institute of Technology, OMR, Chennai, Tamilnadu- - 600119, India
Modalavalasa Divya Indian Assistant Professor, Department of Computer Science and Engineering, Aditya Institute of Technology and Management, Tekkali, Andhra Pradesh-532212, India.
3. Preamble to the description:
The following specification particularly describes the invention and the manner in which it is to be performed.











4. DESCRIPTION

FIELD OF THE INVENTION
This invention focuses on using artificial neural networks (ANNs) to detect and classify fake accounts on social media platforms. It addresses challenges related to online security, user authenticity, and data integrity by leveraging machine learning algorithms to analyze user behavior, interaction patterns, and account characteristics. The aim is to develop a robust, automated detection system that minimizes the presence of fraudulent accounts, thereby enhancing trust and safety within digital communication networks.

BACKGROUND OF THE INVENTION

Artificial Neural Networks for Social Media Fake Account Detection were created to improve digital security and social media authenticity. Social media is essential to modern communication, with billions of users. Due to its popularity, unscrupulous actors have created false accounts to propagate misinformation, manipulate public opinion, launch phishing attacks, commit fraud, and impersonate others. Fake accounts endanger users and platforms. They promote misinformation, cyberbullying, user distrust, spam, and malware. Traditional rule-based algorithms or manual identification have struggled to keep up with bogus account producers' fast-changing strategies. These methods may not detect sophisticated behavioural patterns or adapt to new fraud types, increasing security risks. By emulating human brain function to learn complicated patterns from massive datasets, Artificial Neural Networks (ANNs) may solve this problem. ANNs can identify bogus accounts using user activity logs, profile metadata, interaction patterns, language usage, and other behavioural factors. ANNs can adapt to harmful user techniques because they can learn from dynamic and non-linear data relationships, unlike rule-based systems. This invention follows the trend towards automated security utilising AI. AI-based models, especially neural network-based ones, excel in dynamic contexts like social media because they can learn and adapt. The idea uses deep learning architectures to improve detection accuracy, reduce false positives, and give real-time analysis to make online places safer and more trustworthy. This idea was developed using machine learning, cybersecurity, NLP, and social network analysis research. It is a major advance in the battle against digital fraud, ensuring that social media companies can protect their users and promote pleasant relationships.

SUMMARY OF INVENTION
The idea uses Artificial Neural Networks to identify and remove phoney social media profiles. This system uses ANNs' pattern recognition skills to accurately identify bogus accounts by analysing user behaviour, interaction patterns, account metadata, and content characteristics. The idea uses a deep learning model architecture to understand complicated, non-linear correlations from big datasets to detect sophisticated fraudulent accounts that escape traditional detection approaches. It classifies based on strange posting patterns, quick friend requests, aberrant engagement rates, duplicate profile information, and questionable language usage. The neural network improves its detection capabilities by training on massive labelled datasets of user profiles, adjusting to fraudulent account creators' techniques. This ANN-based detection system uses real-time analysis to identify and remove fraudulent accounts from social media networks, boosting platform security, user trust, and data integrity. The idea improves detection accuracy and reduces false positives, protecting valid accounts. This idea advances social media cybersecurity by delivering an AI-driven, automated, and adaptable solution to a developing problem. It helps digital social networks preserve authenticity by meeting platforms' scalability and real-time processing needs.

BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1: Depicts the Proposed architecture of fake profile identification using ANN.
Fig. 2: Depicts the multiple layers operate on each other to get best match layer and this process continues till no more improvement left.

BRIEF DESCRIPTION OF THE INVENTION
Social media platforms have rapidly expanded into crucial communication conduits since their beginnings, bringing in billions of users all over the world. This growth has occurred since the platforms were first introduced. In spite of the fact that these platforms offer a big number of opportunities for human connection, marketing, and the growth of enterprises, they also create a significant number of challenges in terms of security. The widespread creation of phoney accounts and the establishment of phoney accounts is one of the most widespread problems that has an impact on the internet. Numerous destructive acts, including the dissemination of false information, the execution of phishing attacks, the initiation of scams, the manipulation of public opinion, the theft of identities, and the compromise of users' privacy, are frequently carried out with the assistance of these bogus accounts. When it comes to recognising and removing fake accounts in an effective manner, the traditional rule-based systems and manual detection approaches have been unsuccessful. It is especially important to keep this in mind in light of the fact that the methods that are utilised by people who create false accounts are growing more complicated.
The problem is addressed by this invention, which presents a solution to the issue by utilising Artificial Neural Networks (ANNs), which are designed to emulate human neural networks in order to learn complex patterns from vast datasets. Different from more conventional approaches, artificial neural networks (ANNs) have the ability to comprehend intricate and ever-evolving relationships between variables. This is in contrast to more traditional methods. The fact that they possess this quality makes them particularly well-suited for activities such as the identification of fake accounts. In order to achieve the aims of the innovation, which will be realised through the use of powerful deep learning algorithms, the detection process will be automated, the number of false positives will be reduced, and the real-time response to potential threats will be improved. When it comes to identifying fake accounts on social media platforms, this technology, which is based on neural networks, provides a solution that is not only scalable and adaptive, but also incredibly accurate. The end consequence is that it enhances the confidence and security of the users.
The utilisation of artificial neural networks, specifically deep neural networks (DNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs), is the core of this breakthrough. These neural networks are used to identify fake accounts. As a result of the fact that every type of artificial neural network architecture contributes to a distinct aspect of detection, the system is able to handle a broad variety of input features and data formats without encountering any difficulties. Deep Neural Networks, sometimes referred to as DNNs, are employed for the purpose of evaluating structured data, which includes account metadata (for instance, the date of creation, the number of friends, and the utmost level of completeness of the profile). Through the utilisation of DNNs, it is possible to recognise patterns that are indicative of fraudulent activity on accounts. Some examples of these patterns are recently created profiles that have an exceptionally high engagement rate, an excessive amount of friend requests within a short period of time, or bio material that is repeated multiple times. Convolutional Neural Networks, on the other hand, are employed for the purpose of processing unstructured data such as images and profile pictures. CNNs are used to automate the processing of such data. The ability of these networks to recognise anomalies, such as the frequent use of stock photos or altered visuals, which are indicators of cases of synthetic accounts, is one example. Recurrent Neural Networks (RNNs), and more specifically Long Short-Term Memory (LSTM) networks, are instruments that are utilised for the purpose of analysing temporal patterns in user behaviour. Among the sequential data that these networks analyse are things like the frequency of posting and the histories of interactions. Through the monitoring of peculiar actions, such as high-volume, repetitive messaging, RNNs are able to identify bots or phoney accounts that fail to maintain consistent, organic patterns over time. This can be accomplished by observing the bots or phoney accounts.
In order to input the information into the ANN models, the system collects data from a wide variety of sources. The primary focus of the system is on the characteristics of accounts and the actions of users. There are a few examples of profile metadata that are collected, including the date that the account was created, the ratio of followers to followers, the degree to which the profile is comprehensive, and the number of times that friend requests are issued. In addition to that, it comprises "user activity logs," which include things like the frequency of posting, the volume of messages, and engagement data, as well as "interaction patterns," which include response times, message types, and network consistency. In addition, it includes "engagement data," which may include engagement data. In addition, content analysis is the process of examining text that was written by users in order to find patterns that are indicative of behaviour that is either automated or artificial. It is possible that these patterns consist of language that is repeated, structures that are not natural, or common phrases that are used in spam. The procedure of feature extraction is what is responsible for transforming these raw data into meaningful inputs for the artificial neural network (ANN) models. The Principal Component Analysis (PCA) method is utilised in this process for the purpose of dimensionality reduction. Word embeddings are utilised for text data, while image preprocessing is utilised for profile images.
A large dataset that is tagged and includes both actual and fake accounts is used to train the artificial neural network (ANN)-based detection system. This dataset is used to train the system. The dataset was developed with the purpose of assuring diversity by covering a wide variety of account types, behavioural trends, and tactics for generating fake accounts. This was taken into consideration during the creation of the dataset. The process of preprocessing, which includes correcting missing values, minimising noise, and ensuring that the data is consistent, is an essential component of the training procedure. The artificial neural network (ANN) models are trained through the utilisation of supervised learning techniques, with labelled data providing straightforward indicators of real and fake accounts. ANNs are used to train the models. Loss functions, such as binary cross-entropy, are sometimes used during the training process in order to evaluate the errors that occur in the prediction process. In the process of hyperparameter tuning, techniques such as grid search and random search are utilised. These techniques help in the development of optimal model configurations that are capable of distinguishing between authentic and manufactured accounts. A technique known as k-fold cross-validation is employed in order to ensure that the model is accurate and reliable. In this method, the dataset is segmented into subsets for the purposes of training and validation, which ultimately allows for an evaluation of the model's performance over a variety of data domains.
After being trained and optimised, the artificial neural network (ANN) models are then deployed in a real-time context on social media sites. This occurs after the modelling process has been completed. It is possible for the system to continuously monitor user behaviours and identify fraudulent accounts based on suspicious behaviour because it is integrated with the security infrastructure that is already present on the platform. The system is able to carry out a variety of measures, such as the notification of users of potential hazards when dealing with profiles that are suspected of being fraudulent, the temporary suspension of accounts that are extremely suspicious, and the flagging of accounts for additional evaluation. These are just some of the measures that can be carried out by the system. Through the ability of the artificial neural network (ANN) to process enormous volumes of data in a short period of time, it is possible for real-time detection to take place. The ability to quickly identify and respond to activities involving fake accounts is made possible by this method. The models are able to participate in adaptive learning, which means that they can be retrained on a regular basis with new data in order to keep up with the ever-evolving techniques for building fake accounts. This is possible because the models are able to learn from their own experiences.
The effectiveness of the artificial neural network (ANN)-based detection system is evaluated using standard measures such as accuracy, precision, recall, F1-score, and ROC-AUC (Receiver Operating Characteristic - Area Under Curve). These measurements are utilised in order to assess the effectiveness of the system. While accuracy evaluates the proportion of accounts that have been correctly classified, precision analyses the ratio of true positives to the total number of accounts that have been reported as false. Precision is a measure of precision. It is referred to as recall, and the F1-score is a measure that provides a measure that is intermediate between accuracy and recall. Recall is an evaluation of the system's potential to identify genuine phoney tales, including those that are not immediately clear. In order to determine whether or not the system is capable of distinguishing between genuine and fabricated accounts, the ROC-AUC metric would be utilised. Scores that are higher imply that the system is better able to differentiate between the two different kinds of accounts. Evaluation on a regular basis serves to ensure that the system continues to maintain high levels of accuracy, precision, memory, and effectiveness. This is accomplished through continuous monitoring and adjustments to address any potential defects that may be present in the system.
The invention offers a variety of advantages that are not accessible through the use of traditional methods of detection. The ability of artificial neural networks (ANNs) to comprehend deep links in human activity enables them to achieve higher detection rates than rule-based systems. This is because ANNs are able to perceive more complex patterns of human behaviour. Their ability to achieve high accuracy is a result of this. It is possible for neural networks to continue to be effective over time because of their ability to adjust to new data patterns and false account methods. The plasticity of neural networks brings about the possibility of this happening. Additionally, the system is scalable, which means that it is able to properly accommodate large platforms that have millions of users. Because of this, it is suited for implementation on a worldwide scale. The use of real-time processing makes it possible for the system to readily identify and remove fake accounts in a timely way, thereby protecting the trust of users and the integrity of the platform. Additionally, the technology that is based on artificial neural networks is able to reduce the frequency of false positives by learning complex patterns across a wide range of parameters. Taking this measure helps to ensure that legitimate users are not reported in an inappropriate manner.








, Claims:WE CLAIM
1. The invention claims an automated detection system that utilizes Artificial Neural Networks (ANNs) to accurately identify fake accounts on social media platforms by analyzing user behavior, account characteristics, and interaction patterns in real time.
2. The invention incorporates multiple ANN architectures, including Deep Neural Networks (DNNs) for structured data, Convolutional Neural Networks (CNNs) for image analysis, and Recurrent Neural Networks (RNNs) for sequential behavior monitoring, ensuring comprehensive detection across diverse data types.
3. The invention claims a method for extracting and analyzing key features from user profiles, activity logs, interaction patterns, and content data using advanced preprocessing techniques.
4. The invention enables real-time detection and response to fake account activities through continuous monitoring of user interactions, along with an adaptive learning mechanism that retrains the ANN models periodically to address evolving tactics used by fake account creators.
5. The invention claims a high detection accuracy rate by employing advanced training and optimization techniques, such as supervised learning, hyperparameter tuning, and k-fold cross-validation, to ensure reliable identification of fake accounts with minimal false positives.
6. The invention is designed to be scalable, capable of integrating with large-scale social media platforms to handle high volumes of user data, making it suitable for global deployment across various digital environments.
7. The invention provides an enhanced cybersecurity measure for social media platforms by accurately distinguishing between genuine and fake accounts.

Dated this the 26th October 2024.


Senthil Kumar B
Agent for the applicant
IN/PA-1549

Documents