Tell me again, why should i talk to strangers?

In the Internet of Things, it is not straightforward to decide which devices can interact, or how. Under a unique label we are hiding devices with a large variety of capabilities and means of communications. It is therefore not realistic to assume that a single security framework will suffice for all cases. Yet, it is critical to make sure that communications are properly authorized, and possibly secured, to avoid a variety of issues including hijacking and eavesdropping. In this paper, we identify different perspectives under which devices can communicate and their specific requirements, as embedded, peered and independent. In each case, we look at the nature of the communications and the key issues of identification, authorization, and protection. Rather than looking at those classical problems from a device perspective on top of an open, end-to-end network, we propose instead that a platform perspective is more beneficial, and help integrate security issues into the design, rather than an afterthought.