E-mail has traditionally been regarded as the most powerful medium in online social networks, where users can discuss, connect, and share links with other online social media users. In particular, Twitter, in particular, has been determined to be the most popular social network that serves as the best communication channel for its users to share current news, ideas, thoughts, comments, and beliefs with other online social media users. Despite the efforts put in to combat spam operations on the online social network, Twitter spam has a new type of functionality that is limited to 140 characters. It is not only the major cause of annoyance for day-to-day users, but also responsible for the majority of computer security issues that cost billions of dollars in terms of productivity losses. In this paper, we propose a Multi-Objective Genetic Algorithm and a CNN-based Deep Learning Architectural Scheme (MOGA–CNN–DLAS) for the predominant Twitter spam detection process. The experimental details and results discussions of the proposed MOGA-CNN-DLAS are evaluated in terms of accuracy, precision, recall, F-Score, RMSE, and MAE by varying the ratio of training data under the utilization of three real datasets, such as the Twitter 100k dataset and the ASU dataset.