AI-based intelligent sensing detection of cybersecurity threats using multimodal sensor data in smart devices.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Internet of Things (IoT)

Background:

• Internet of Things (IoT) systems present unique cyber-physical vulnerabilities.
• Traditional network-based intrusion detection struggles with on-device sensor-generated malicious activities.

Purpose of the Study:

• To introduce a multimodal sensing architecture for detecting cyber-attacks in IoT devices.
• To leverage heterogeneous sensor data (acceleration, gyroscope, microphone, temperature) for enhanced threat identification.

Main Methods:

• Developed a hybrid deep learning architecture combining Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Transformers.
• Implemented feature fusion and spatial-temporal interaction analysis across sensor modalities.
• Evaluated the framework on a custom multimodal dataset and public datasets (CICIDS-2017, IoT-23).

Main Results:

• Achieved a high Area Under the Curve (AUC) of 0.96 and an F1-score of 0.94.
• Demonstrated low inference latency of 23 ms on edge hardware, confirming real-time deployability.
• Validated the effectiveness of multimodal deep learning for cyber-physical threat detection.

Conclusions:

• Multimodal deep learning offers a powerful and scalable method for cyber-physical threat detection in IoT.
• The proposed architecture effectively addresses the limitations of traditional intrusion detection systems in IoT environments.
• The framework is suitable for real-time deployment in resource-constrained IoT settings.