 Hi, I'm Peter, and I want to talk to you about LaMoll, which is the dynamic and fluid MPC protocol for dishonest majority. In secure multipart computation, a set of parties who each have some private input want to jointly compute some function F across all of their inputs. Of course, they want to do this while preserving privacy of their inputs. In this talk, I'm going to focus on the dishonest majority setting with the majority of these parties may be corrupted by the adversary potentially maliciously. Typically an MPC, we consider a fixed set of parties which doesn't change over time. But today I want to consider dynamic participants. The main motivation for this is flexibility. So if the parties are running a very long and expensive computation, they might not want to commit to being online for the whole time, or they may want to drop out of the way through. Another benefit though can actually be security, moving the participants over time creates a moving target for the adversary, which may be harder to corrupt. This has taken advantage of, for instance, in the Yoso model for MPC, where each party only actually is online to send a single message in the whole protocol. So our work is based on the Fluid MPC model by Chaudhri et al from crypto 2020. So in there, the protocol proceeds in time periods called epochs, where each epoch runs within a single fixed committee of parties. And they work in the honest majority setting, so they assume that each committee in one epoch has an honest majority of the parties. And the epochs may proceed in a series of rounds of interaction amongst that committee. And then at the end of the epoch, the parties will hand over their state to the next committee before going offline. So in our work, we extend this to the dishonest majority setting. And one challenge we have here is that most practical dishonest majority protocols for MPC actually rely on a pre-processing phase, which is a kind of a more expensive phase used to generate correlated randomness done ahead of time and independently of the inputs of the computation. So to model this in Le Mans, we allow all of the parties to be online in the pre-processing phase. And then after that, we move to the online phase where the computation is divided into these epochs with changing committees. And since we're in dishonest majority, the security, we only require that there's at least one honest party in each committee within any epoch. So in Le Mans, we give two different variants of the speeds protocol for multiparty computation in this model. The first of these, we call dynamic speeds, where we relax the online phase to only actually have a single epoch with one committee. And this can, of course, be any subset of the parties who took part in the pre-processing phase doesn't need to be fixed until the online phase starts. So this is more practical, a little more restricted with only having one committee online. In the fluid speeds protocol, the other hand, we have full fluidity, meaning that we allow evolving online committees in the online phase, changing from one epoch to another. The protocol can even be maximally fluid, meaning that each epoch only consists of a single round of communication before the party can go offline and move on to the next committee. So one thing that ties together both of these protocols is the pre-processing phase. And since the pre-processing involves all of the parties, we design a new kind of universal pre-processing protocol which generates correlated randomness that can be later and adaptively used by any subset of parties later on. And we do this by building on recent advances in pseudo random correlation generators. So another benefit of the dynamic speeds protocol is that we actually get something very practical. And the pre-processing phase has the lowest communication of any speeds like pre-processing protocol that was previously known. This even holds if you look at only protocols without the dynamic feature. So that's all I have time for now. But if you want to read more, you can look at our paper on e-print or check out the talk at crypto next week.