AutoLiquid-FD: Next-Gen Liquid Neural Network for Fraudulent Transaction Detection

As cyber threats continue to grow in complexity and scale, real-time detection and rapid response have become essential for safeguarding network environments. This paper explores the transformative role of artificial intelligence (AI) in advancing next-generation information security through automated, real-time fraud detection. The main objective of this study was to present a model to detect fraudulent transaction using AutoLiquid-FD (ALFD) (Combination of Autoencoder and Liquid Neural Network (LNN)) algorithm. In this novel hybrid algorithm, Autoencoder for creating a lower-dimensional representation of the data and LNN for detecting fraud transaction which could potentially collect more complex patterns in sequential or time-series data. This model is tested with new pre-processed transactions, if the fraud probability value is higher than the threshold value (0.5) that is consider as fraud transaction, otherwise legitimate transaction. This model’s performance metrics are accuracy 99.65, precision 92.89, recall 89.28 and F1-score 90.48.