Now, in the age of smart manufacturing, while embedding IoT devices into industrial processes enhances operational efficiency. It also exposes them to a host of potential dangers including malware that can specifically target these sensor networks. This work contributes to better solving a demand for malware detection by means of comprehensive performance evaluation studies among multiple machine learners, where it tries comparison between these tools and usage in practice. The main goal was to evaluate the performance of k-nearest neighbor (KNN), Random forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) in anomaly detection using sensor data (capacitive and inductive proximity sensors; magnetometer Hall effect sensors; encoders.) The technique used to pre-process a dataset of 2300 sensor readings per sensor type for normal and abnormal signals. These data were split into training and testing sets, with models trained using accuracy, precision, recall, F1-score as the evaluation metric. The results obtained shows that ANN reached an accuracy of 97.86%, with higher precision and recall. F1-score again debiting on their good performance in predicting the variation of sensor readings due to malware. KNN and RF also performed well, with accuracies of 94.56% and 92.30%, respectively, demonstrating their effectiveness in pattern recognition and classification. Although SVM showed a slightly lower accuracy of 88.34%, it provided valuable insights into complex data patterns.