EFFICIENT INTERNET COMMUNITY ANALYSIS USING INTEGRATED ADABOOST-CATBOOST FOR FAKE ACCOUNT DETECTION

Abstract

Detecting fraudulent accounts in Internet communities is a difficult undertaking due to fraudsters' clever methods and the large amount of user data involved. Concerns about security and authenticity in online social networks (OSNs) have grown as a result of their widespread incorporation into daily life. The rise in popularity of OSNs has attracted unscrupulous actors looking to exploit personal data and spread misleading information. While academics have investigated different ways for detecting aberrant behaviors and bogus accounts, current systems frequently suffer from inadequate feature selection and suboptimal classification algorithms. In this research, we propose ADB-CB, a unique algorithm aimed to improve the detection of phony Instagram accounts. To increase accuracy and reliability, we use four feature selection and dimension reduction algorithms, including AdaBoost, CatBoost, and ExtraTree classifiers. The experimental findings show that our strategy works well for identifying between authentic and phony Instagram profiles.

Specification

Description:The HybridDetect Algorithm is a comprehensive framework designed to detect fake Instagram
profiles effectively. It begins with data collection, gathering user attributes such as names,
bios, follower counts, and activity logs. The collected data undergoes preprocessing to handle
missing values and inconsistencies, followed by feature selection to identify relevant attributes
indicative of fake accounts. Dimensionality reduction techniques are then applied to simplify
the data while retaining essential information. , Claims:1. The HybridDetect Algorithm employs a structured approach to detect fake Instagram
profiles, utilizing comprehensive feature selection (102) to identify attributes indicative
of fraudulent accounts.
2. The algorithm applies dimensionality reduction techniques (103) to enhance
computational efficiency while preserving essential data information.
3. The HybridDetect Algorithm integrates AdaBoost, CatBoost, and Extra Trees
classifiers (104) within an ensemble learning framework to improve classification
performance.
4. The final step accurately classifies (105) Instagram profiles as real or fake based on the
ensemble model's predictions, ensuring reliable detection of fraudulent accounts.

Documents