With recommendation systems for anomaly detection across diverse areas such as healthcare, pharmaceuticals, automobiles, and finance, the advancement of Machine Learning (ML) has impacted every facet of human existence. The widespread use of ML has created an urgent need to protect the privacy of many stakeholders of ML algorithms, including data owners, model owners, and inference users. Privacy-Preserving Machine Learning (PPML), through an active area of research, lacks either technique with ML accuracy or performance. This research, CoInMPro-DP, focuses on solving ML inference inefficiency by proposing an efficient and confidential inference. It uses Secure Multi-party Computation (SMPC) techniques to protect model parameters and applies Differential Privacy (DP) at the inference level to protect the data owner’s privacy. The suggested method can increase the privacy of ML stakeholders, including data owners, model owners, and users of inference, without sacrificing accuracy and incurring a small performance cost. The Python implementation used SyMPC and PyDP open-source libraries for SMPC and DP-based functions and ran experiments on the Boston Housing dataset, another open-source dataset. Exhaustive experimentation was carried out on Azure Data Science VMs with Ubuntu OS. The results proved that privacy enhancement for ML stakeholders had no substantial impact on accuracy, even with a high privacy budget on a bigger scale. To enhance privacy, when the secure node counts are increased in the SMPC cluster, the findings showed a linear influence on inference performance.