Enhanced SqueezeNet model for detecting IoT-Bot attacks: A comprehensive approach.

Area of Science:

• Cybersecurity
• Machine Learning
• Deep Learning

Background:

• The rapid expansion of Internet of Things (IoT) devices has amplified cyber threats, particularly botnet attacks targeting network security.
• Existing machine learning (ML) and deep learning (DL) methods for detecting these threats face challenges with accuracy and computational demands, limiting their real-time application in resource-limited IoT settings.

Purpose of the Study:

• To develop an advanced detection framework that enhances the accuracy and computational efficiency of identifying cyber threats in IoT environments.
• To address the limitations of current ML/DL approaches for real-time intrusion detection within the constraints of IoT systems.

Main Methods:

• An improved SqueezeNet model was integrated with a Deep Convolutional Neural Network (DCNN) and an optimized stochastic mixed Lp layer.
• Data pre-processing involved min-max normalization for consistent scaling and improved model learning.
• Feature extraction and classification were performed using the integrated SqueezeNet-DCNN architecture.

Main Results:

• The proposed model achieved a high classification accuracy of 0.97.
• A significantly reduced false positive rate of 0.054 was observed.
• Experimental results demonstrated superior performance compared to established techniques like Bi-GRU, CNN, PolyNet, and LinkNet.

Conclusions:

• The enhanced SqueezeNet-DCNN framework offers a computationally efficient and accurate solution for real-time botnet attack detection in IoT environments.
• The proposed model represents a significant advancement in securing IoT networks against sophisticated cyber threats.
• The findings suggest the model's suitability for practical deployment in resource-constrained IoT security applications.