<p>In the existing Scenario, threats are getting larger as a consistent cybersecurity problem, which attacks computer systems, handheld devices, and ubiquitous networks. The current generation of highly critical malware has also surpassed the traditional detection methods in various aspects such as accuracy, flexibility and resilience to the original attack methods. The given research is a systematic literature review of threat detection aiming to examine the literature published since 2015 to 2025 to assess the current status of the field of research and categorize the main issues or possible directions of further research that warrant further research. A new taxonomic system of discovery modes depending on deep learning and machine learning is proposed, which is viewed by datasets, feature extaction systems and algorithmic categorizers. Moreover, it gives some discussion of experimental bias that has a major influence on the performance of malware detection and defines critical measures of effectiveness to evaluate the effectiveness, and certain undefined research problems. This paper introduces possibilities of improving detection method and creates insights on plausible solutions and future research directions. Lastly, the current paper discusses the anomalies that may be experienced in digital twin and how to identify.</p>