The security of computer networks is of paramount importance in today’s digital age, where network intrusion attempts are becoming more sophisticated and frequent. Intrusion detection systems (IDS) which were established long ago are used extensively in the industry but are also limited in their ability to detect modern intrusion techniques that are often obfuscated and hidden in network traffic. As a result, there is a growing need for advanced techniques in network security that can effectively identify and mitigate such attacks.This study proposes a novel activation function TReLU, which is implemented with a hybrid model of CNN and BiLSTM. This function edges out the currently used activation functions for this use case. Additionally, to reduce the training time, parallel programming in Python is leveraged to accelerate the computation of the model. We have used three benchmark IDS datasets to evaluate our proposed model with TReLU. Compared to the previous works of other authors, the experimental results depict that our model outperforms or slightly edges them and various other state of the art models.