A Resilient Property Authentication Framework with Decentralized Records and Intelligent Fraud Analytics

Abstract

The rapid expansion of decentralized digital systems has intensified the demand for intelligent and secure fraud detection mechanisms capable of handling complex transaction data. With the increasing adoption of blockchain platforms such as Ethereum, ensuring transaction authenticity and preventing fraud has become a significant challenge. Traditional verification methods rely heavily on manual processes, making them inefficient for large-scale datasets and leading to issues such as low accuracy, poor scalability, high processing time, and limited transparency. To address these challenges, this study proposes a hybrid fraud detection framework that integrates multiple Machine Learning (ML) algorithms, including Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGB), and Light Gradient Boosting Machine (LGBM), along with a Deep Learning (DL) model named LandGuardNet based on a 2D Convolutional Neural Network (CNN2D). Feature selection is performed using Analysis of Variance (ANOVA), and normalization techniques are applied to ensure consistent data representation. All models are trained and evaluated in a Jupyter Notebook environment using standard performance metrics. Experimental results indicate that the CNN2D-based LandGuardNet model outperforms other models in detecting complex and non-linear fraud patterns and is selected for deployment. The system is implemented using the Django Web Framework, providing role-based access for buyers and sellers. Additionally, secure storage is ensured using IPFS, while transaction records and prediction results are stored on the Ethereum blockchain via smart contracts, ensuring immutability, transparency, and traceability.