Pengembangan Sistem Deteksi Anomali Berbasis Artificial Intelligence untuk Meningkatkan Keamanan Informasi dan Ketahanan Sistem Digital

Abstract

Purpose: This study develops a conceptual architecture for artificial-intelligence-based anomaly detection to improve information security and the operational resilience of digital systems amid evolving cyber threats.

Design/Methodology: A critical methodological review integrates statistical baselines, supervised and unsupervised learning, autoencoders, sequence models, and risk-oriented incident response. Model design, data pipelines, evaluation criteria, governance, and deployment are analyzed without asserting any unperformed experiment.

Findings: Anomaly detection can identify deviations not captured by static signatures, but effectiveness depends on representative data, realistic temporal validation, calibrated alerts, adversarial robustness, and responses that do not disrupt essential services.

Practical Implications: The proposed layered workflow supports security operations teams in prioritizing alerts, measuring false positives, identifying drift, preserving audit trails, and coordinating detection with recovery and continuity planning.

Originality/Value: The synthesis connects predictive anomaly scoring with resilient system operation through a transparent framework encompassing detection, human verification, adaptive response, and continuous model assurance.