Paper Details

Name: Jitendra Mina, Gadil Ferooz, Shahrukh Khan, Chandrashekhar Ajay, Lokendra Pratap & Sarthak Gondwal
Country: India

Year: 2026
Volume: Volume No: 13, January, Year: 2026 (Special Issue)
Page Number: 787-793
DOI: https://doi.org/10.5281/zenodo.19129892

ABSTRACT—
Real-time anomaly detection from streaming data is vital in applications like cybersecurity, fraud analytics, and industrial monitoring, where timely detection of suspicious patterns is imperative. This paper proposes a scalable and effective
anomaly detection framework using Apache Flink’s distributed stream processing feature to handle high-velocity data with low latency. By combining machine learning methods with Flink’s event-driven architecture, the system proposed here efficiently detects anomalies in real-time with high throughput and computational efficiency. The method is tested on benchmark datasets, showing its flexibility to dynamic streaming environments and its capability to detect anomalies with high accuracy. Performance indicators like detection rate, processing time, and resource usage are examined to confirm the efficacy of the framework, rendering it a solid solution for real-time anomaly detection in massive data streams.

Index Terms—Real-time anomaly detection, streaming data, Apache Flink, machine learning, distributed processing, eventdriven architecture, scalability, fraud detection, industrial monitoring,