JoWUA

Abstract

The growth of online platforms has triggered the growing necessity of effective detection systems to deal with threats such as fake news, spam, and unauthorized users. Existing models have serious problems, such as limited task coverage, limited generalization across diverse datasets, overfitting, and the inability to address multi-domain threats within a single framework. As a way to overcome these obstacles, we would suggest the Multi-Tiered Architecture for Spam Data Detection (MTA-SD), a solution that will incorporate AI-based methods, including machine learning, deep learning, and natural language processing, into a multi-layered detection architecture. The system is able to address more than fake news, spam, and unauthorized user detection, but by augmenting digital platforms' security, it supports the objective of SDG 9: Industry, Innovation, and Infrastructure, by providing safe, robust, and innovative infrastructures. The architecture addresses three major detection tasks simultaneously, including Fake News Detection (FND), Spam or Malicious Data Detection (MDD), and Unauthorized User Detection (UUD). Using the hybrid systems of BERT with textual data, GPT-2 with contextual insights, and XGBoost for spam classification, the proposed system is guaranteed to achieve high accuracy across different types of inputs. The results of the evaluation indicate that the MTA-SD model is better than the previous solutions since it has an accuracy of 99.90, precision of 99.95, recall of 99.96, and an F1 score of 99.99% in various datasets, including ISOT, LIAR, IFND, and Malicious Webpages. This capability of the architecture to combine various detection activities, make use of sophisticated feature engineering, and evolve with new threats with the continuous learning of new features makes it a scalable and robust system to achieve real-time detection. This model not only addresses the constraints of the traditional methodologies but also offers a multifaceted, flexible, and very precise system for solving the current issues of digital security.