Deep learning technology: enabling safe communication via the internet of things
IntroductionThe Internet of Things (IoT) is a new technology that connects billions of devices. Despite offering many advantages, the diversified architecture and wide connectivity of IoT make it vulnerable to various cyberattacks, potentially leading to data breaches and financial loss. Preventing...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-02-01
|
| Series: | Frontiers in Communications and Networks |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/frcmn.2025.1416845/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | IntroductionThe Internet of Things (IoT) is a new technology that connects billions of devices. Despite offering many advantages, the diversified architecture and wide connectivity of IoT make it vulnerable to various cyberattacks, potentially leading to data breaches and financial loss. Preventing such attacks on the IoT ecosystem is essential to ensuring its security.MethodsThis paper introduces a software-defined network (SDN)-enabled solution for vulnerability discovery in IoT systems, leveraging deep learning. Specifically, the Cuda-deep neural network (Cu-DNN), Cuda-bidirectional long short-term memory (Cu-BLSTM), and Cuda-gated recurrent unit (Cu-DNNGRU) classifiers are utilized for effective threat detection. The approach includes a 10-fold cross-validation process to ensure the impartiality of the findings. The most recent publicly available CICIDS2021 dataset was used to train the hybrid model.ResultsThe proposed method achieves an impressive recall rate of 99.96% and an accuracy of 99.87%, demonstrating its effectiveness. The hybrid model was also compared to benchmark classifiers, including Cuda-Deep Neural Network, Cuda-Gated Recurrent Unit, and long short-term memory (Cu-DNNLSTM and Cu-GRULSTM).DiscussionOur proposed technique outperforms existing classifiers based on various evaluation criteria such as F1-score, speed efficiency, accuracy, and precision. This shows the strength of the approach in threat detection and highlights the potential of combining SDN with deep learning for IoT vulnerability assessment. |
|---|---|
| ISSN: | 2673-530X |