Fully decentralized learning is a setting in which each peer in a P2P network trains a machine learning model with the help of the other peers. Each peer acts as a model manager by periodically sending her current model to other peers, who answer by returning model updates they compute on their private data. This creates a tension among privacy, accuracy and security. The privacy risk is that model updates returned by a peer can leak some of the peer’s private data. Unfortunately, distorting model updates to protect privacy works against the accuracy of the trained model. On the other hand, aggregating the updates of several peers and then sending the aggregate to the model manager may preserve privacy but it goes against security, because the model manager cannot filter out individual bad updates. Also, peers are autonomous and hence it cannot be taken for granted that they will honestly supply model updates to help the model manager train her model. To reconcile accuracy, privacy and security, we present a fully decentralized learning protocol such that: (i) it allows perfectly accurate individual updates to be returned by peers to the model manager in a privacy-preserving manner; (ii) it is co-utile by design, that is, it incentivizes rational peers to follow the protocol without deviating. The latter feature discourages rational attacks that might compromise security and it also deters free riding, thereby ensuring the sustainability of the protocol.

Multiparty computation (MPC) consists in several parties engaging in joint computation in such a way that each party's input and output remain private to that party. Whereas MPC protocols for specific computations have existed since the 1980s, only recently general-purpose compilers have been developed to allow MPC on arbitrary functions. Yet, using today's MPC compilers requires substantial programming effort and skill on the user's side, among other things because nearly all compilers translate the code of the computation into a Boolean or arithmetic circuit. In particular, the circuit representation requires unrolling loops and recursive calls, which forces programmers to (often manually) define loop bounds and hardly use recursion. We present an approach allowing MPC on an arbitrary computation expressed as ordinary code with all functionalities that does not need to be translated into a circuit. Our notion of input and output privacy is predicated on unlinkability. Our method leverages co-utile computation outsourcing using anonymous channels via decentralized reputation, makes a minimalistic use of cryptography and does not require participants to be honest-but-curious: it works as long as participants are rational (self-interested), which may include rationally malicious peers (who become attackers if this is advantageous to them). We present example applications, including e-voting. Our empirical work shows that reputation captures well the behavior of peers and ensures that parties with high reputation obtain correct results.

The decentralized nature of federated learning, that often leverages the power of edge devices, makes it vulnerable to attacks against privacy and security. The privacy risk for a peer is that the model update she computes on her private data may, when sent to the model manager, leak information on those private data. Even more obvious are security attacks, whereby one or several malicious peers return wrong model updates in order to disrupt the learning process and lead to a wrong model being learned. In this article, we build a federated learning framework that offers privacy to the participating peers as well as security against the Byzantine and poisoning attacks. Our framework consists of several protocols that provide strong privacy to the participating peers via unlinkable anonymity and that are rationally sustainable based on the co-utility property. In other words, no rational party is interested in deviating from the proposed protocols. We leverage the notion of co-utility to build a decentralized co-utile reputation management system that provides incentives for parties to adhere to the protocols. Unlike privacy protection via differential privacy, our approach preserves the values of model updates and, hence, the accuracy of plain federated learning; unlike privacy protection via update aggregation, our approach preserves the ability to detect bad model updates while substantially reducing the computational overhead compared to methods based on homomorphic encryption.

Broadcast encryption (BE) schemes allow a sender to securely broadcast to any subset of members but require a trusted party to distribute decryption keys. Group key agreement (GKA) protocols enable a group of members to negotiate a common encryption key via open networks so that only the group members can decrypt the ciphertexts encrypted under the shared encryption key, but a sender cannot exclude any particular member from decrypting the ciphertexts. In this paper, we bridge these two notions with a hybrid primitive referred to as contributory broadcast encryption (ConBE). In this new primitive, a group of members negotiate a common public encryption key while each member holds a decryption key. A sender seeing the public group encryption key can limit the decryption to a subset of members of his choice. Following this model, we propose a ConBE scheme with short ciphertexts. The scheme is proven to be fully collusion-resistant under the decision n-Bilinear Diffie-Hellman Exponentiation (BDHE) assumption in the standard model. Of independent interest, we present a new BE scheme that is aggregatable. The aggregatability property is shown to be useful to construct advanced protocols.

The problem of efficiently and securely broadcasting to a remote cooperative group occurs in many newly emerging networks. A major challenge in devising such systems is to overcome the obstacles of the potentially limited communication from the group to the sender, the unavailability of a fully trusted key generation center, and the dynamics of the sender. The existing key management paradigms cannot deal with these challenges effectively. In this paper, we circumvent these obstacles and close this gap by proposing a novel key management paradigm. The new paradigm is a hybrid of traditional broadcast encryption and group key agreement. In such a system, each member maintains a single public/secret key pair. Upon seeing the public keys of the members, a remote sender can securely broadcast to any intended subgroup chosen in an ad hoc way. Following this model, we instantiate a scheme that is proven secure in the standard model. Even if all the nonintended members collude, they cannot extract any useful information from the transmitted messages. After the public group encryption key is extracted, both the computation overhead and the communication cost are independent of the group size. Furthermore, our scheme facilitates simple yet efficient member deletion/addition and flexible rekeying strategies. Its strong security against collusion, its constant overhead, and its implementation friendliness without relying on a fully trusted authority render our protocol a very promising solution to many applications.