Cyberattacks in the digital world have been continuous against its individual, business, and government segments. Along with it, the threat space is dynamic and changing at a very rapid pace, where securing a network's infrastructure is of prime concern these days. In this paper, a new detection mechanism of these cyberattacks will be proposed using an ensemble of machine learning models based on voting classifiers. In this paper, various ensemble methods based on AdaBoost, KNN, gradient boosting, logistic regression, and support vector machines has recommended to incorporate for improved accuracy and reliability in the detection of attacks. A dense dataset is drawn from the network traffic report in this study in the composition of normal combined with attack categories. The extracted features combine packet size, protocol type, and connection duration, which are merged into training the voting classifier. This is an ensemble learning model that merges a lot of weak learners into one robust classifier. The reasons for choosing this regression version of the voting classifier are its intrinsic ability to work with continuous target variables, hence perfectly fitting cyberattack classification. Other than computational efficiency for real-time applications, classification performance is measured using evaluation metrices: accuracy, positive predictive value (precision), sensitivity (recall), and the F1-score.