A Convolutional Neural Network-Based Deep Learning Framework for Automated Cyberattack Detection in IoT Applications

With the combination of diverse IoT technologiesand no global standards imposed, the IoT has furtheropened up revolutionary, innovative applications yethas also presented new and complex securitychallenges. Restrictions are needed to safeguard IoTapplications since cyber threats are increasing daily.Artificial Intelligence (AI) has made it possible tosolve many real-world issues. This study introducesthe Learning-based Cyberattack Detection system(LbCADF). This deep learning-based systememploys a CNN-based model with enhancedsensitivity for autonomously detecting cyberattacksin IoT settings. The framework successfullydistinguishes between benign and malevolent trafficflows. We add feature selection and hyperparametertweaking to improve training quality to our proposedsystem, Enhanced CNN for Attack Detection andClassification (ECNN-ADC). To preventoverfitting, an early stopping criterion is applied.This work has been evaluated on the benchmarkdataset UNSW-NB15. At the same time, theempirical results indicate that ECNN-ADC achievesthe highest detection accuracy (95%) compared tovarious of the latest models (MLP, baseline CNN).