AI-Powered Anomaly Detection in Zero Trust Environments: A Comprehensive Review of Methods and Evaluation
Year: 2024
Type: article
Abstract: Advanced cyberattacks challenge modern cybersecurity. New threats render perimeter security useless. The potential ZTNA promotes "never trust, always verify" for access control. ZTNA finds network irregularities. AI and ML may improve ZTNA anomaly detection, says this study.ZTNA anomaly detection ML algorithms are theoretically designed first. This research trains algorithms utilizing pre-labeled datasets of typical and atypical network traffic patterns using Supervised Learning. Labeled data patterns allow SVMs and Random Forests to classify new network activity as normal or abnormal. ZTNAs with rapidly expanding attack pathways may label data poorly.Unsupervised learning excels with unlabeled data. These algorithms predict unlabeled network traffic behavior. Baselines show anomalies. PCA and K-Means discover abnormalities. Without labeled data, unsupervised learning may miss malevolent outliers.RL investigates ZTNA anomalies. For trial-and-error learning, RL algorithms use incentives and penalties. ZTNA RL agents can monitor network traffic and block suspicious connections based on security system input. While RL can adjust dynamically, training is computationally expensive and setup-dependent.Evaluate any installed ML algorithm. The paper attacks ZTNA anomaly detection metrics. Precision and recall, which measure the ratio of correctly recognized anomalies to true abnormalities, are crucial for determining the system's ability to identify normal and abnormal actions. The harmonic mean of Precision and Recall, F1-score, shows system equilibrium. System anomalous sensitivity and false positive rate are measured by DR and FAR.ML algorithms and performance measures are used to select and evaluate ZTNA anomaly detection solutions. This method allows security experts secure networks with AI and ML.
Source: Nanotechnology Perceptions
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Keywords: Zero Trust Network Access, anomaly detection, network traffic analysis, supervised learning, unsupervised learning, reinforcement learning, support vector machine, random forest, principal component analysis, K-means clustering, precision and recall, false positive rate, false rejection rate, network security
Open Access status: diamond
OpenAlex ID: W7217153999
