In recent communication technology, number of attacks and threats caused by intruders, who interrupt secure transmission of confidential data for their own use. Here, the intruders steal information of authorized users which causes trouble in the area of medical, industries, business, and so on. Therefore, several deep learning approaches are emerged to detect the intruders over the authorized network. Due to a number of limitations including imbalanced data, computational complexity, and other security issues, the previous methods do not provide accurate detection. This work concentrated on proposing a new approach, the Twin Adam optimizer-based deep learning model (TA-DL) for effective intrusion detection (ID). This research takes the deep convolutional neural network (CNN) as learning model, which utilizes the deep features to increase the efficiency of the model thereby increasing security for communication. Moreover, the model adopts the proposed Twin Adam optimizer, ensemble with root mean squared propagation (RMSProp), and Nestero-accelerated Adam, which is developed to tune the parameters of the TA-DL model. Therefore, the model outperformed high ID and the performance is evaluated using accuracy, sensitivity, and specificity metrics. The TA-DL achieved 98.02% of accuracy compared to other conventional methods.