As unmanned aerial vehicles (UAVs) are rapidly being adopted in various industries, there is a need to formulate concepts for safe operability. In this paper, the problem of identifying malicious UAVs is tackled by employing a multi-criteria decision making (MCDM) approach which is based on supervised machine learning. The approach looks to enhance the efficiency and precision of UAVs classification by measuring several performance parameters such as the fall rate, delivery rate, and loss rate, as well as residual energy, among others. MCDM helps in giving a decision on the classification of a given UAV in the dataset as either normal or malignant. After this, the features of the dataset are standardized to undertake inter-comparison. The user-trained classifiers include Random Forest, Decision trees, Naïve Bayes, and Support Vector Machine models where cross-validation sequences are employed to test for overfitting. The results support the effectiveness of Random Forest with classification accuracy reaching 99.9% in execution time of about 0.45 s. These results also indicate the possibility of improving the operational security of Flying Ad-Hoc Network (FANETs) by enhancing the capability of malicious nodes identification. This framework employs multi criteria decision making (MCDM) enabled by machine learning to tackle the problem of malicious identification clearly.