Kazi Tuhin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Kazi Tuhin has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Niropam Das
International American University
2 shared publications - Partha Chakraborty
International American University
2 shared publications - Talha Bin Ansar
Katz School of Science and Health, Yeshiva University
2 shared publications - Ahmed Shan-A-Alahi
University of The Potomac
1 shared publication - Habiba Rahman
International American University
1 shared publication - Harleen Kaur
Westcliff University
1 shared publication
Latest Publications
- 2 August 2026
- Published by Springer Nature
- p. 317-331
Resilient MIS Design Using Cryptographic and Cyber-Security Techniques
- 2 August 2026
- Published by Springer Nature
- p. 317-331
Resilient MIS Design Using Cryptographic and Cyber-Security Techniques
- 1 February 2026
- Published by Springer Nature
- p. 488-504
A Privacy-Preserving Federated Intrusion Detection System Leveraging Secure Multi-Party Computation Across Collaborative Cloud Tenants
- 1 February 2026
- Published by Springer Nature
- p. 488-504
A Privacy-Preserving Federated Intrusion Detection System Leveraging Secure Multi-Party Computation Across Collaborative Cloud Tenants
- 13 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Continual Learning Pipelines for Detecting Insider Threats Across Corporate Cybersecurity Infrastructures
By virtue of the fact that authorized individuals are able to circumvent traditional barriers, insider attacks continue to be a persistent worry for organizational cybersecurity. Different user behaviors and complex connection patterns can cause static detection systems to miss potential dangers and slow down the discovery process. An adaptive insider threat detection system and a real-time adaptive security framework are both presented in this study. SentriDeep-RL, which is a sentinel based on deep learning and reinforcement learning, is also shown. At the same time that graph neural networks (GNNs) simulate user-asset interactions, deep learning models represent behavioral characteristics. Within the SentriDeep-RL framework, real-time anomaly detection frameworks are able to identify dangers in an instant, and adaptive thresholding that is based on reinforcement learning is able to make dynamic decisions about threats that are unknown. Changing insider threat patterns can be identified by the system without the need for retraining because it monitors user behaviors, access records, and network telemetry. Gradual model modifications and temporal attention are utilized in the SentriDeep-RL approach. In early threat detection, SentriDeep-RL outperforms static models by 12-18% on multisource corporate datasets that imitate both planned and incidental insider acts. This results in a reduction in the number of false positives while keeping the same level of computing efficiency! Adaptive learning methods and idea drift detection are two techniques that can be utilized to obtain improved performance in changing business environments. Through the utilization of deep learning, GNN-based relational modeling, reinforcement learning, and real-time anomaly detection, this platform will be able to identify insider threats in a proactive, scalable, and resilient manner. Last but not least, SentriDeepRL leverages technology to address concerns regarding company cybersecurity, reaction time, as well as operational and reputational issues.
- 13 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Continual Learning Pipelines for Detecting Insider Threats Across Corporate Cybersecurity Infrastructures
By virtue of the fact that authorized individuals are able to circumvent traditional barriers, insider attacks continue to be a persistent worry for organizational cybersecurity. Different user behaviors and complex connection patterns can cause static detection systems to miss potential dangers and slow down the discovery process. An adaptive insider threat detection system and a real-time adaptive security framework are both presented in this study. SentriDeep-RL, which is a sentinel based on deep learning and reinforcement learning, is also shown. At the same time that graph neural networks (GNNs) simulate user-asset interactions, deep learning models represent behavioral characteristics. Within the SentriDeep-RL framework, real-time anomaly detection frameworks are able to identify dangers in an instant, and adaptive thresholding that is based on reinforcement learning is able to make dynamic decisions about threats that are unknown. Changing insider threat patterns can be identified by the system without the need for retraining because it monitors user behaviors, access records, and network telemetry. Gradual model modifications and temporal attention are utilized in the SentriDeep-RL approach. In early threat detection, SentriDeep-RL outperforms static models by 12-18% on multisource corporate datasets that imitate both planned and incidental insider acts. This results in a reduction in the number of false positives while keeping the same level of computing efficiency! Adaptive learning methods and idea drift detection are two techniques that can be utilized to obtain improved performance in changing business environments. Through the utilization of deep learning, GNN-based relational modeling, reinforcement learning, and real-time anomaly detection, this platform will be able to identify insider threats in a proactive, scalable, and resilient manner. Last but not least, SentriDeepRL leverages technology to address concerns regarding company cybersecurity, reaction time, as well as operational and reputational issues.