Abstract: Breach detection helps in identifying unauthorized access or suspicious activities in a network system. It detects threats at an early stage before they cause serious damage, protecting sensitive information like personal data, financial records, and confidential files. Breach detection systems provide real-time alerts, allowing security teams to take quick action and prevent further harm. However, most existing breach detection systems face challenges like false alarms, slow detection speed, and inability to detect new attacks. Some systems also struggle to detect intrusions in encrypted data and consume high system resources. These limitations affect the accuracy and performance of the detection system. To overcome these issues, the breach detection system is implemented using Temporal Fusion Transformers (TFT), which analyses time-based patterns in network traffic to detect intrusions accurately. The current study incorporates the Simargyl2022 dataset to enhance the quality of our results and analyses, which contains both normal network traffic and malicious attack data, making it suitable for evaluating detection performance. The system achieved 95.40 accuracy, with a recall of 95.40, precision of 91.01, and an F1-score of 93.15, showing its high efficiency in detecting breaches. The outcomes of this study have significant implications for network security, providing valuable insights for practitioners and researchers working towards building robust and intelligent breach detection systems.
Key Words: Breach Detection, Temporal Fusion Transformer (TFT), Cybersecurity, Anomaly Detection, Time-Series Forecasting, Temporal Dependencies, Multi-Headed Attention, Gating Layers, Scalable Systems, Advanced Deep Learning.
|
DOI:
10.17148/IJARCCE.2025.14381
