Machine learning (ML) is driving advances in industries ranging from healthcare and finance to transportation and infrastructure, and therefore security and resiliency are a must. ML-based Cyber-Physical systems (CPS) such as autonomous vehicles, medical diagnostics, and smart energy grids are highly susceptible to sophisticated cyberattacks. Adversarial attacks capable of modifying data to disrupt the system are elusive, making traditional security methods, like encryption and intrusion detection, insufficient to protect against them. Moreover, protecting sensitive personal data in ML applications presents unique privacy challenges, especially in healthcare. Measuring impacts to encrypted models while at the same time, existing propositions such as SecureML refine the privacy of such models by encrypting the underlying data, this leads to unwanted delays in processing time due to the compromise in both performance and security. To alleviate these issues, in this paper, we propose ParSecureML, a novel GPU-based framework that builds upon secure multi-party computation (MPC) to ensure a better balance between data security, operational efficiency, and privacy in ML applications. With the power of ParSecureML, we are one step closer to developing resilient systems that can continue to operate even when challenged, enabling secure and efficient ML as we move on to increasing connectivity across the globe.