The problem of data security in 5G-IoT network has been well studied. The support of rapid communication provided by the network has been used by various sectors. The security threat has been identified in the same frequency as it supports the communication. Towards handling such intrusion threats, there exist numerous techniques which address specific problem and consider only limited factors of communication. This inclined their performance against security towards different threats. To address this issue, an efficient Real-time Feature Impact Optimization (RFIO) based DNN model (RFIO-DNN) is presented in this article. The method uses multiple standard data sets for the evaluation of the methods in restricting different threats. To start with, the data sets are merged and normalized to a single entity using Feature Level Fuzzy Normalizer. Second the method applies Feature Impact Optimization (RFIO) algorithm towards feature selection. Using the features selected, the method trains the deep neural network. At the test phase, the neurons of the network compute Feature Level Trust (FLT) and Transmission Level Trust Weight (TLTW). Using these values, the output layer neuron computes Multi Constraint Trust Weight (MCTW) according to the IoT devices present in the transmission route. Incoming data are classified with MCTW towards intrusion detection. © 2024 Elsevier B.V., All rights reserved.