DEEP LEARNING–DRIVEN REAL-TIME ANOMALY DETECTION FOR TIME-SERIES DATA IN CLOUD ENVIRONMENTS

Abstract

Purpose: The present study proposes to explore the impact of using real-time anomaly detection using deep learning in cloud-based system, specifically analysing its role in the detection of anomalies in time-series data, cloud security, operational performance, scalability, and implementation challenges. The study also considers intelligent anomaly detection technologies' impact on future cloud architectures from a strategic perspective. Design/Methodology/Approach: The research design used was quantitative with a structured questionnaire that was sent to 280 professionals from the cloud computing, cybersecurity, data analytics, machine learning and related technology industry. A total of 30 items were measured using a five-point Likert scale distributed over six major constructs. The data were analyzed using descriptive statistical data, reliability analysis (Cronbach's Alpha) and chi-square testing using SPSS. Findings: Reliability analysis showed that the reliability of the questionnaire was very high, and the Cronbach's Alpha values ranged from 0.87 to 0.93 and the overall reliability coefficient was 0.91. The results found high levels of awareness and usage of deep learning technologies (M = 4.16) and the highest level of agreement was for cloud performance (M = 4.28) and security enhancement (M = 4.28). Other factors such as effectiveness of anomaly detection (M = 4.27), future directions and strategic impact (M = 4.31), and scalability and real-time processing (M = 4.20) were also highly rated. Both respondents highly agreed that deep learning is beneficial for real-time threat detection, operational efficiency, cloud reliability, and proactive risk management. But the issues of computational cost, model interpretability, privacy and availability of skilled professionals continued to be big hurdles. The results of all Chi-square showed p < 0.01 which revealed high level of consensus among participants. Originality/Value: The outcomes of this study give empirical evidence for the systems developed for the detection of anomalies using deep-learning models adopted, which were able to be effective, scalable, and potentially applicable for future intelligent monitoring of the cloud. The outcomes can be valuable for cloud service providers, cybersecurity professionals, researchers, and organizational decision makers looking to leverage the latest AI tools and techniques to improve cloud resilience, security, and operational efficiencies.

Keywords : Deep Learning, Anomaly Detection, Cloud Computing, Time-Series Data, Cybersecurity, Real-Time Monitoring.