Mutida: A Rights Management Protocol for Distributed Storage Systems Without Fully Trusted Nodes

Abstract

Several distributed storage solutions that do not rely on a central server have been proposed over the last few years. Most of them are deployed on public networks on the internet. However, these solutions often do not provide a mechanism for access rights to enable the users to control who can access a specific file or piece of data. In this article, we propose Mutida (from the Latin word "Aditum" meaning "access"), a protocol that allows the owner of a file to delegate access rights to another user. This access right can then be delegated to a computing node to process the piece of data. The mechanism relies on the encryption of the data, public key/value pair storage to register the access control list and on a function executed locally by the nodes to compute the decryption key. After presenting the mechanism, its advantages and limitations, we show that the proposed mechanism has similar functionalities to Wave, an authorization framework with transitive delegation. However, Wave does not require fully trusted nodes. We implement our approach in a Java software program and evaluate it on the Grid'5000 testbed. We compare our approach to an approach based on a protocol relying on Shamir key reconstruction, which provides similar features.is that there are no servers that we can rely on to be in charge of the protocol for rights management. Additionally, anonymity is a key point since peers are communicating directly and additional layers need to be in place to guarantee it. [19]. In this paper, we propose Mutida (from the Latin word "Aditum" meaning "access" and written from right to left), a protocol that focuses on enabling users to manage access rights in a network of the second category.In this paper, we consider the files stored on the Interplanetary File System [11] (IPFS), which is a storage solution that relies on a BitTorrent-like [35] protocol to exchange the files between the nodes and on a Kademlia [39] Distributed Hash Table (DHT) to locate the different pieces of data in the network. The essential characteristic of IPFS is that files are immutable and cannot be modified once they are written.Using a DHT forwards all location requests through different nodes. Therefore, any connected node is involved in the routing of requests and can determine the identifiers of popular files [26]. Additionally, since the storage solution does not manage any permission and because IPFS does not provide any encryption mechanism to protect user privacy [47], every user can access all the files stored in it if the content identifier (CID) is known.The consequence of this is that any node in the IPFS network can observe the DHT requests and access the corresponding files. This illustrates and justifies the need to manage access permissions in such a network. With Mutida, all users will still be able to find the files in the network, but only the permitted users will be able to decrypt the content.Because of the use of Merkle trees [40] and the use of a root hash as a file identifier, users of IPFS do not need to tru...