The Internet of Things has made our lives easier to live and more convenient. However, the risks of cyber-attacks, especially within the kernels of the IoT, have increased manifold. A strong security system is required to ensure that the devices are safe from any threat. Therefore, proposing an augmented pattern evaluation model that makes use of Deep Dyna Q and VARMA GRU-based predictive analysis to be able to provide additional embedded security features to IoT kernels. Altogether, model that composes of three components: feature extraction, model training, and prediction operations. At the first component, extraction of the relevant features from the IoT kernel in order to create a feature vector for this process is carried out. The feature vector is further utilized to train the Deep Dyna Q algorithm, a reinforcement learning approach which learns to decide under the maximization of some reward signal. Here the use the second module with the VARMA GRU-based predictive analysis algorithm to predict the future state of the IoT kernel based on the current state and actions taken by the Deep Dyna Q algorithm. The VARMA GRU algorithm implements a VARMA with the advantages of a GRU model and thus provides forecasts accurately. In the final component, assess the predicted state of the IoT kernel with a set of predefined security rules. If any of these rules are broken by the predicted state, the system acts accordingly to reduce the possible threats. This will be the model that would consider an all-encompassing security of IoT kernels by harnessing the power of various algorithms. The Deep Dyna Q ensures that the system will make intelligent decisions in real-time, while the VARMA GRU adds accuracy with its predictive analysis algorithm, hence making this an augmented pattern evaluation model that would rise to the ever-increasing security challenges of IoT devices and deployments.