In this paper, we present a novel approach for epidemic information dissemination, also known as gossiping, in sensor networks. By construction, our protocol exhibits three interesting properties. First, the routing of messages follows a topology with characteristics similar to that of scale-free graphs, which have been shown to be particularly efficient in terms of power consumption, when gossiping in wireless networks. Then, our protocol is fully decentralized, meaning that each sensor works independently, i.e., no coordination is necessary among sensors. Finally, it is stateless, meaning that no history information is necessary. In particular, no neighborhood knowledge is necessary to execute the protocol. Intuitively, our decentralized and stateless protocol dynamically modulates its transmission power according to a power-law distribution. That is, our approach allows sensors to save power with no negative impact, neither on the time needed for reaching all the sensors, nor on the number of messages sent.