Secure Internet of Things (IoTs) have evolved into a requirement for electronic healthcare systems. In most cases, health images contain sensitive information about patients that must be protected. Traditional encryption cannot be directly applied to image data due to restrictions in digital data attributes. Additionally, patients may lose the confidentiality of their data when private images are transmitted via a network. Thus, multimedia Artificial Intelligence and image processing are applied to build improved secure IoTs. To guarantee accurate and privately protected e-health services, a secure lightweight key frame extraction approach is essential. Additionally, when taking into account the limitations of real-time e-health systems, it can be challenging to establish a satisfactory degree of security in an economical manner. An encryption scheme that contain a hashing version of the Blum Blum Shub (BBS) generator, namely Hash-BBS (HBBS) is suggested and built to achieve a high grade of integrity and confidentiality in transmission data of COVID-19 CT-images for patients. Also, an AI technique is applied for COVID-19 testing such as adopted a convolutional neural network. Evaluation showed that the proposed framework outperformed alternative security and transfer learning methodologies in secure prediction. Therefore, it can be used to reliably transmit CT-images for COVID-19 patients while meeting strict security and prediction benchmarks.