Comparative Analysis of Fake Account Detection using Machine Learning Algorithms Dr A Padmavathi Dept. of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Chennai, India a padmavathi@ch.amrita.edu K.B.Vaisshnavi Dept. of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Chennai,India kbvaisshnavi@gmail.com Abstract—With the rise of Social Media platforms and new applications, the Rapid Expansion of fake accounts has become an important concern, posing threats to security, privacy and trustworthiness. In response, this research explores the appli- cation of machine learning techniques for the detection and reduction of fake accounts. By utilizing datasets containing, user behavior patterns, network characteristics and various parameters, we explore the efficiency of various machine learning algorithms in selective genuine users from false ones. This study includes attribute manipulation, model training and evaluation techniques made for the unique challenges of fake account detection. We examine the effects of different features, such as interaction measures, language patterns and network structures on the output of detection models. Through systematic testing and validation, we identify effective methods for accurately identifying fake accounts in broad virtual spaces. Our results to the progress of information security efforts and the improvement of user trust in online platforms. Index Terms—Fake account detection, Machine learning, So- cial media, Network analysis, Algorithms I. I NTRODUCTION In the constantly growing field of social media, the rise of fake accounts has come up as a strong challenge, threatening the validity and security of online communities.The universal nature of fake accounts is often created with evil motive, destroys trust, compromises user privacy and enables various forms of online manipulation and fraud. As social media platforms continue to grow in effect and user base, the universality of fake accounts presents an important issue that requires immediate and effective detection and mitigation strategies. Fake accounts can be used to circulate misinformation and manipulate public opinion which can create artificial trends and follower counts which are very essential for marketing, public relations and policy-making. For businesses, fake accounts can lead to distorted data analytics, affecting decision making and development of strategies. For individual users, these fake accounts often lead to breaches of personal data, emotional distress and financial loss. The manipulation of social media activities by fake accounts are also a threat to democratic events such as elections through coordinated disinformation campaigns.Hence, there is an urgent need for strong and effective ways to detect and reduce the spread of fraudulent groups. This research project is for addressing this critical issue by using the power of machine learning especially Artificial Neural Networks (ANN) for the detection of fake accounts [1].This strategy involves constructing a sophisticated model that consists of various sets of user accounts attributes which are the key indicators in the process of classification of accounts. The study focuses on utilizing a dataset having different attributes taken from user profiles, including the information of profile pictures, the length of usernames,follower and followings counts and the number of posts.By analyzing attributes, the model can capture anomalies and patterns that are characteristics of fake accounts. The classification is structured around binary leading, where accounts are marked as ’0’ for fake and ’1’ for genuine.By training the model on a robust dataset, it can identify differences and relationships between the input features that may not be evident through simpler methods. The main objective of this research paper is to develop a classification model of detection which is capable of accurately differentiating the genuine and fake accounts based on the given attributes. To accomplish this, we use ANN architecture made for the features of the dataset. The model training involved utilizing parameters using techniques like ’adam’ optimizer ’categorical crossentropy’ loss function,with accuracy as the performance metric. This training process which is measured over multiple epochs, aims to increase the model’s predictive capabilities. Major steps in the procedure including data preparation,feature selection, model training and evaluation. In data preparation, we clean and adjust the dataset to ensure the working. Feature selection involves important attributes that are expressive of account realistic or fake. This model is trained using the training dataset with the aim of learning complex patterns and connections with the data. Evaluation of ANN model’s efficiency is done using testing dataset allowing for proper analysis of its effectiveness in detecting fake accounts. Performance metrics such as accuracy and precision are used to evaluate the model’s efficiency and measure its strength in practical situations.This research also