The substitution box (s-box) serves as one of the key elements of cryptography responsible for uncertainty and nonlinear properties. These boxes must have a variety of security qualities in order to protect a cipher against various attacks, including side-channel attacks. Designing a 16*16 s-box that is both cryptographically secure and energy efficient is a difficult task. Thus, in order to automate the process of determination and verification of dynamic power efficiency of s-boxes, a method of supervised machine learning approach has been used here. Additionally, utilizing the outcomes of supervised learning, in order to construct s-boxes that are both cryptographically secure and low-power, we propose a deterministic model that may be used in an optimization strategy to estimate the dynamic strength of an s-box. A machine learning- approach has been integrated to automate and categorize the power efficiency. It is evident that there is a 4% increase in the accuracy of power prediction of s-box.