Emerging decentralized storage services such as Storj and Filecoin show promise as a new paradigm for data outsourcing. These services tie cryptocurrency to personal storage resources and leverage blockchain technology to ensure data integrity in distributed networks. Compared to current cloud storage, they are expected to be more scalable, cost effective, and secure. In addition to the features above, strong guarantees of data privacy are seriously desired due to today's prevalent data leak and abuse incidents. However, simply using end-to-end encryption limits the search capability and thus will degrade the user experience. In this paper, we propose an encrypted decentralized storage architecture that can support trustworthy and private keyword search functions. We start from searchable encryption to achieve search on encrypted data. Yet, only adopting this primitive is not sufficient to address particular threats in our target decentralized service model. Service peers would maliciously return incorrect results, while user peers would fraudulently refuse to pay service fees. To resolve those threats, we devise specific secure data addition and keyword search protocols to enable client-side verifiability and blockchain based fair judgments on the search results. For practical considerations, we integrate an efficient dynamic searchable encryption scheme to our protocols as an instantiation to lower the blockchain overhead. Our security and performance analysis indicates the advance of the proposed architecture.