A Model for Detecting Illegitimate Access in Financial Databases Using Deep Neural Networks
This research explores Deep Neural Networks (DNN) as advanced classifiers for detecting illegitimate access to financial databases. Aim: The aim of this research is to build a deep learning model capable of identifying illegitimate access to financial databases The dataset was obtained from the Kaggle Machine Learning Repository (Phishing Dataset) and verified against PhishTank data for accuracy. It contains 10,000 labeled URLs (5,000 phishing and 5,000 legitimate). The implementation was conducted using Anaconda3, Jupyter Notebook with Tensorflow, Keras, Pandas, Numpy and Flask framework. The proposed model achieved accuracy of 99.76%, complemented by a precision of 98.69%, a recall of 99.64%, and an F1-score of 99.50%. Additionally, the ROC-AUC score stands at 0.99.
Download Paper