A smart weight scale, or smart scale, is an Internet of Things (IoT)-based scale that can measure metrics other than body weight using various sensors and send the information to the cloud. Meanwhile, the problem is that a person’s weight is considered personally identifiable information (PII) and needs to be preserved to comply with general data protection regulations (GDPR). Our research aim is to use K-Means for anonymization so that a privacy-preserving smart body scale becomes GDPR-compliant. The first step is to form a novel privacy-preserving smart body scale framework. We obtained the cardiovascular disease dataset containing personal weights from Kaggle. We apply random noise perturbation and k-means clustering for anonymization. We apply cardiovascular disease classification using gradient boosting. Finally, we compared the performance of the three anonymization methods with several metrics, including information loss, entropy, and distortion. Test results show that our elbow method shows that the optimum number of clusters for body weight is six. This number has passed the k-anonymity assessment. Furthermore, comparisons show that the k-means generalization performs better than noise perturbation with distortion, information loss, and entropy values 71.1, 0.001, and 15.6, respectively.