<p>Data security in electronic health records (EHR) remains a critical concern, with existing methods employing session, service, feature, rule, and profile-based encryption and access restriction models. However, these methods often fall short in achieving optimal security performance, leading to diminished trust among data owners. To address this challenge, we propose an Adaptive Feature-Centric Polynomial Data Security Model (AFPDSM) capable of efficiently enforcing security measures on diverse datasets. Based on data taxonomy analysis, the AFPDSM divides EHR data aspects into discrete groups according to their significance. The technique adapts the ciphertext produced by polynomial functions for data encryption and creates a blockchain based on feature values. Additionally, the model implements a Healthy Trust Access Restriction scheme to prevent malicious access. Adoption of the AFPDSM can result in up to 98% improvement in security performance and up to 97% improvement in access restriction performance. When compared to existing methods like HCA-ECC, EHRCHAIN, IBE-DLM, and SPAKE, our approach showcases significant improvements. Specifically, we observed a remarkable 98% enhancement in throughput. Additionally, our method exhibits reduced computational complexity, with an average encryption time of 101&#xa0;ms and an average decryption time of 104&#xa0;ms. Moreover, the network overhead is notably reduced by 21% compared to conventional methods. These findings underscore the superior performance and efficiency of our proposed approach in handling electronic health records (EHRs).</p>