Scholarlink

The accelerated growth of ransomware and malware presents substantial obstacles to conventional cybersecurity strategies, which often depend on static signatures and struggle to detect complex, evolving cyber threats. This research study emerging ransomware and malware detection by integrating deep learning methods into behavior based intelligent firewall systems, utilizing Zigbee networks for secure communication. A novel hybrid framework, called the Behavioral-Attention Graph Neural Network (BA-GNN), was introduced by combining Graph Neural Networks (GNNs) with attention mechanisms to analyze real time behavioral patterns and identify zero day ransomware cyberattacks. This security framework incorporates Adaptive Behavioral Signature Profiling (ABSP) and Deep Behavioral Sequence Mapping (DBSM) to adapt to emerging cyber threats in real time. Moreover, a reinforcement learning based ensemble model is proposed, which dynamically selects and weightings classifiers, such as Naïve Bayes, Random Forest, and Artificial Neural Network system, to enhance detection precision and reduce false positives rate. The experimental results demonstrate that the GN-BiLSTM method achieved 99.99% accuracy in ransomware detection, whereas the BA-Graph Neural Networks framework improved the detection rates for obfuscated and zero-day variants by 5% compared to existing methods. Further, the incorporation of API feature engineering and behavior graph models enhanced the detection resilience, achieving 98.5% accuracy in identifying abnormal network behaviors. This research contributes to a proactive cyber defense mechanism against evolving cyber threats, mitigating the cyber risks associated with data breaches and financial data losses. Future directions include optimizing computational efficiency, enabling continuous model updates, and expanding these techniques to 5G/6G enabled IoT landscapes. This systematic approach holds significant potential for securing smart cities and critical infrastructure, and fostering a reliable and flexible digital framework.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Learning and Reinforcement Learning for Proactive Ransomware Detection: A Behavioral-Attention Graph Neural Network Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sai Kiranmai,
  • P. Senthilkumar

摘要

The accelerated growth of ransomware and malware presents substantial obstacles to conventional cybersecurity strategies, which often depend on static signatures and struggle to detect complex, evolving cyber threats. This research study emerging ransomware and malware detection by integrating deep learning methods into behavior based intelligent firewall systems, utilizing Zigbee networks for secure communication. A novel hybrid framework, called the Behavioral-Attention Graph Neural Network (BA-GNN), was introduced by combining Graph Neural Networks (GNNs) with attention mechanisms to analyze real time behavioral patterns and identify zero day ransomware cyberattacks. This security framework incorporates Adaptive Behavioral Signature Profiling (ABSP) and Deep Behavioral Sequence Mapping (DBSM) to adapt to emerging cyber threats in real time. Moreover, a reinforcement learning based ensemble model is proposed, which dynamically selects and weightings classifiers, such as Naïve Bayes, Random Forest, and Artificial Neural Network system, to enhance detection precision and reduce false positives rate. The experimental results demonstrate that the GN-BiLSTM method achieved 99.99% accuracy in ransomware detection, whereas the BA-Graph Neural Networks framework improved the detection rates for obfuscated and zero-day variants by 5% compared to existing methods. Further, the incorporation of API feature engineering and behavior graph models enhanced the detection resilience, achieving 98.5% accuracy in identifying abnormal network behaviors. This research contributes to a proactive cyber defense mechanism against evolving cyber threats, mitigating the cyber risks associated with data breaches and financial data losses. Future directions include optimizing computational efficiency, enabling continuous model updates, and expanding these techniques to 5G/6G enabled IoT landscapes. This systematic approach holds significant potential for securing smart cities and critical infrastructure, and fostering a reliable and flexible digital framework.