The use of covert channel techniques has increased the capacity to carry out dangerous and undetectable attacks. Traditional security procedures cannot identify them because they utilize methods not meant to transmit information. A covert channel type that is difficult to identify, reduce the impact of, or eradicate is a packet-length covert channel. This covert channel makes use of differences in network packet lengths to send secret messages. Recent studies have emphasized the advantages of using machine learning techniques to identify covert channel attacks. As a result, this work offered an effective ensemble classification model to find these kinds of assaults. Three machine learning techniques make up the ensemble model, which serves as our model's primary classifiers. These classifiers consist of Support Vector Machine (SVM), Random Forest (RF), and Naive Bayes (NB) (SVM). The output of the proposed ensemble classifier was produced by combining the primary classifiers' outputs using the logistic regression (LR) classifier which is served as a meta classifier. Our proposed ensemble model performed well, according to the results. It surpasses all single classification algorithms by achieving a considerable accuracy rate to detect such type of covert channel attacks.