Insecure smart home IoT network is growing in number and size. It puts smart home network security and users privacy at risk and impacts the security of the rest of Internet community (e.g. botnets). The diversity, lack of update, and nature of IoT market mean that we can't rely on patching. Researchers have explored enforcing network policy to enhance the security of IoT as it doesn't require amendment on the end-devices. However, the rule of IoT authorities and end-users in securing smart home IoT has not been investigated. Therefore, this research explores the development of a security architecture for smart home IoT that enforce IoT authorities and end-users collaborative security policy. This will utilise software-defined networking (SDN) for dynamic policy control and machine learning techniques to address define policy in an automatic manner. I will address the problem concerning define security policy language, policy conflict, automatic policy definition using machine learning.