The chapter has proposed a security framework for sensory tags on the Internet of Things (IoT) using intelligent Radio Frequency (RF) pattern analysis with machine learning to identify malicious tags. The RF Identification (RFID) system is an integral part of sensory tag based IoT applications. Due to limited computational capabilities on tags, traditional security primitives are not feasible for sensory tags, while there are existing lightweight security techniques that can be used to protect expensive RFID sensory tags, like active tags. However, less expensive tags, like passive tags, are incapable of using most of the existing lightweight security techniques. In addition, low-cost printed chipless sensory tags are incapable of using any existing security primitives at all due to no available computational and storage capabilities on board. Hence, implementing a holistic security provision to a hybrid sensory tag based IoT eco-system becomes complex or infeasible. The proposed security framework uses transmitted RF signals to counter security attacks, like malware injection, repudiation, and counterfeit, that are usually designed to infiltrated IoT system using malicious tag to address the security needs of all categories of sensory tags in the IoT. The proposed security framework extract features from RF data to prepare a fingerprint of the tag for its identity in association with other tags in the network, which is then used in machine learning for classification. It only needs to use computational and storage power of the reader to free the tags from computation and storage burden. The experiment on the proposed model shows that it can identify malicious tags with a high accuracy which is validated against several other machine learning techniques.