Resilient Infrastructure through Autonomous Machine Learning and Cloud-Native Observability for Threat Detection

Abstract

Modern digital infrastructure is increasingly distributed across cloud platforms, containers, microservices, edge devices, application programming interfaces, and dynamically provisioned computing resources. This complexity expands the attack surface while making conventional monitoring and rule-based threat detection less effective against rapidly changing and previously unseen attacks. This paper examines how autonomous machine learning combined with cloud-native observability can strengthen infrastructure resilience and threat detection. Autonomous machine learning refers to systems capable of continuously learning from operational and security telemetry, identifying deviations from established patterns, adapting detection models, and supporting automated responses with limited human intervention. Cloud-native observability integrates metrics, logs, traces, events, network flows, and security signals to provide a contextual representation of infrastructure behaviour. The proposed approach considers a unified architecture in which telemetry is collected continuously, processed through scalable cloud-native pipelines, analysed using machine-learning techniques, and correlated to identify anomalous or malicious behaviour. The methodology incorporates data preprocessing, feature engineering, unsupervised and supervised learning, continuous model evaluation, and automated response mechanisms. Particular attention is given to explainability, false-positive reduction, model drift, scalability, privacy, and adversarial manipulation. The study argues that resilient infrastructure requires not only accurate threat detection but also continuous visibility, adaptive learning, fault tolerance, and controlled automation