.png){kind=icon}
Return
Efficientnetv2-RegNet: an effective deep learning framework for secure SDN based IOT network
.png){kind=icon}
PRE
AI
DOI：10.1007/s10586-024-04498-0.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Traditional network administration required manual programming of routing policies and related parameters on specific routers and switches, which was expensive. Therefore, software-defined networking (SDN) technology has been introduced, which has boosted flexibility and decreased hardware development costs by centralizing network management. Since intrusion detection is vital in the SDN environment, this centralized architecture makes information security vulnerable to network threats. To evaluate and recognize these attacks, many researchers have recently adopted cutting-edge approaches like machine learning. However, most of these methods are not very accurate and scalable. To address this issue, this paper proposes an EfficientNetV2-RegNet-based effective deep learning technique. It effectively extracted the network features and classified the intrusions in SDN-based IoT (Internet of Things). Afterwards, an effective mitigation process was performed by a remote SDN controller to mitigate the assaults and reconfigure the network resources for trusted network hosts. Furthermore, the Conditional Generative Adversarial Network (CGAN) based data augmentation approach efficiently tackles the data imbalance issue. The most recent realistic datasets, named InSDN and IoT-23, were utilized to train and assess the presented framework to validate its efficiency. The results of the experiments demonstrated that the suggested system surpassed competitors in identifying various attack types and achieved 99.53 and 99.56% accuracy for IoT-23 and InSDN datasets, correspondingly.
Keywords:
EfficientNetV2
SDN
Intrusion detection'
RegNet
Mitigation
Cycle GAN
Journal
C
IF:
4.1
Papers:
5.1K
Citations:
7.5K
Organization
Cited Papers
.png){kind=icon}
ShareSaveShareSave
View more