The process of ensuring the performance and action of the edge devices to access the resources is secured by authenticating and authorizing the devices. As the devices are increasing due to its huge demand, the unauthorized access in the network is increasing that affects the security of the edge device data. The existing models tried to detect the authorized and unauthorized access in the network for reducing the vulnerable activities but failed to handle the dimensionalities of the features that reduced the security of the system. The embedded bidirectional encoder representations from transformers with regularized random forest (EBERT-RRF) model is developed to overcome the existing problem for the detection of authenticating and authorizing the edge devices. The embedded bidirectional encoder representations from transformers (EBERT) model extracted the significant information from the edge device data that increased the training ability of the model by minimizing the overfitting problem. The regularized random forest (RRF) classifier was used for the detection of authentication and authorization of edge devices. The developed EBERT-RRF model has obtained better results of f1-score of 0.859, accuracy of 0.925, recall of 0.853, and precision of 0.867 compared to the existing extreme gradient boosting (XGB).