Deep Learning–Enabled Honeypots: An ANN-Based Approach for Advanced Cyber Threat Analysis

A cyber threat refers to an illegal activity intended to breach the confidentiality and integrity of computer systems and data. Examples of cyber threats include malware, phishing, unauthorized computer access, and denial-of-service attacks. It has made the traditionally relied-upon Intrusion Detection System based on signature systems obsolete and incapable of providing the much-needed protection against hacking incidents. Keeping this concern in mind, we propose an innovative concept for Deep Learning-Enabled Honeypot Cyber Attack Systems based on Honeypot-Based Artificial Neural Network Systems to facilitate efficient intelligence for effective analysis of cyber attacks. Experimental evaluation of the proposals is conducted using the CIC-IDS-2017 dataset for the attack scenario under consideration throughout the simulation. An intensive data preprocessing technique is employed to address high dimensionality, noisy features, and imbalance. The proposed HP-ANN model is systematically compared to several machine learning and deep learning baselines, namely Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). From the results, the proposed HP-ANN model significantly outperforms all baselines in terms of accuracy, precision, recall, and F1-score, achieving 1.00 and a very near-perfect ROC-AUC. Besides, the model's stability and fast learning ability are undeniable, as demonstrated by additional analyses of convergence trends and confusion matrices.