We consider consensus protocols in the model that is most commonly considered for use in state machine replication, as initiated by Dwork-Lynch-Stockmeyer, then by Castro-Liskov in 1999 with "PBFT." Such protocols guarantee, assuming n players out of which t < n/3 are maliciously corrupted, that the honest players output the same valid value within a finite number of messages, after the (unknown) point in time where both: the network becomes synchronous, and a designated player (the leader) is honest. The state of the art (Hotstuff, PODC'19), achieves linear communication complexity, but at the cost of additional latency, due to one more round-trip with the leader. Furthermore, it relies on constant-size threshold signatures schemes (TSS), for which all prior-known constructions require a costly interactive (or trusted) setup. We remove all of these limitations. The communication bottleneck of PBFT lies in the subprotocol, denoted as "view change," in which the leader forwards 2t+1 signed messages to each player. Then, each player checks that these 2t+1 messages satisfy some predicate, which we denote "non-supermajority''. We replace this with a responsive subprotocol, with linear communication complexity, that enables players to check this predicate. Its construction is elementary, since it requires only black box use of any TSS. In the full version of our paper \citemalicious2 we achieve three things. Firstly, we further optimize this subprotocol from succinct arguments of many signed messages, which we instantiate from Attema-Cramer-Rambaud \cite[2021-3-9 version]ACR20. As an introduction to these methods, we discuss here the simplest case, which is the construction in \citeACR20 of the first logarithmic-sized TSS with transparent setup. Second, we also address another complexity challenge pointed in Hotstuff, namely, that protocols with fast termination in favorable runs, have so far quadratic complexity, due to an even more complex view change. Third, we enable halting in finite time with (amortized) linear complexity, which was an unsolved question so far when external validity is required.

We apply multiparty computation (MPC) techniques to show, given a database that is secret-shared among multiple mutually distrustful parties, how the parties may obliviously construct a decision tree based on the secret data. We consider data with continuous attributes (i.e., coming from a large domain), and develop a secure version of a learning algorithm similar to the C4.5 or CART algorithms. Previous MPC-based work only focused on decision tree learning with discrete attributes (De Hoogh et al. 2014). Our starting point is to apply an existing generic MPC protocol to a standard decision tree learning algorithm, which we then optimize in several ways. We exploit the fact that even if we allow the data to have continuous values, which a priori might require fixed or floating point representations, the output of the tree learning algorithm only depends on the relative ordering of the data. By obliviously sorting the data we reduce the number of comparisons needed per node to O(N log2N) from the naive O(N2), where N is the number of training records in the dataset, thus making the algorithm feasible for larger datasets. This does however introduce a problem when duplicate values occur in the dataset, but we manage to overcome this problem with a relatively cheap subprotocol. We show a procedure to convert a sorting network into a permutation network of smaller complexity, resulting in a round complexity of O(log N) per layer in the tree. We implement our algorithm in the MP-SPDZ framework and benchmark our implementation for both passive and active three-party computation using arithmetic modulo 264. We apply our implementation to a large scale medical dataset of ≈ 290 000 rows using random forests, and thus demonstrate practical feasibility of using MPC for privacy-preserving machine learning based on decision trees for large datasets.

Abstract We consider secure integer division within a secret-sharing based secure multi-party computation framework, where the dividend is secret-shared, but the divisor is privately known to a single party. We mention various applications where this situation arises. We give a solution within the passive security model, and extend this to the active model, achieving a complexity linear in the input bit length. We benchmark both solutions using the well-known MP-SPDZ framework in a cloud environment. Our integer division protocol with a private divisor clearly outperforms the secret divisor solution, both in runtime and communication complexity.

At CRYPTO 2018, Cramer et al. introduced a secret-sharing based protocol called SPD \(\mathbb {Z}_{2^k}\) that allows for secure multiparty computation (MPC) in the dishonest majority setting over the ring of integers modulo \(2^k\) , thus solving a long-standing open question in MPC about secure computation over rings in this setting. In this paper we study this problem in the information-theoretic scenario. More specifically, we ask the following question: Can we obtain information-theoretic MPC protocols that work over rings with comparable efficiency to corresponding protocols over fields? We answer this question in the affirmative by presenting an efficient protocol for robust Secure Multiparty Computation over \(\mathbb {Z}/p^{k}\mathbb {Z}\) (for any prime p and positive integer k ) that is perfectly secure against active adversaries corrupting a fraction of at most 1/3 players, and a robust protocol that is statistically secure against an active adversary corrupting a fraction of at most 1/2 players.