In this era, the use of Internet of Things (IoT) is growing exponentially. This positive development also raises questions about security of IoT. In the I.T industry, many vendors are producing products that are IP-enabled and can easily be configured by just plug and play. This ease is creating heighten security problems as proper identification and authentication are required for each IP-enabled device. In this study, we extend the work of IoT-Sentinel which has been already reported in the literature. In this work, a security mechanism that identifies devices as well as constrains communication of vulnerable devices has been designed and developed. In this paper, we have been able to optimize the earlier work by demonstrating how to identify devices that could not be identified or declared vulnerable with high accuracy using Machine learning. With Machine learning algorithms, we have been able to classify devices of which Random Forest provided the best performance in terms of accuracy than others. The outcome of our approach is more promising and has added a new optimal level of resilience to security.