 Hey So I'm really I'm going to talk about some gen work with hotek and Daniel that we call like an extension evaluation So a standard problem at script. Oh is to have a general chair say tall that wants to distribute socks among Sox among attendees The problem is that we're not happy with just picking socks We want each of us to have the best circle available So after years of research at IBM they managed to produce some algorithm the best possible to pitch The best of four each of us. It turns out that the algorithm is pretty weird first the Yeah, the input to the algorithm is the sorry is the dating history of the attendees So it's pretty weird, but it turned out that this was the best option possible and also another problem was that The algorithm is actually extremely complicated. So it actually requires IBM supercomputers to run it So only talk and I can run it So first she puts at the registration page She asked every attendee to put the dating history, but for some reason people didn't want to do that So as cryptographer, how can we help? So we introduce a new primitive that we call iconic function evaluation Where we have some private data x and we want to to let tall compute f of x on our data So using like a new function evaluation tall can Compress the huge algorithm into some short digest that can be seen as a public key for the function and Your regular crypto attendee will be able to use this public key to generate a safer text of the Of its dating history after that how can recover the cipher text and using her knowledge of the circuit Will be able to recover the best socks so she can put the other so that's nice But more formally what happens is that? We want pretty much or Primitive to only to let tall learn about ffx and certainly means that she only Learns ffx and not the whole dating history and For the iconic part as our name suggests we want the cipher text to be succinct in the sense that Encryption time should be smaller than the Set of distance to our computers. We are not at IBM And we manage to construct such a primitive from LW more seriously Application I'm pretty excited about is on MPC where we have and parties with input X1 up to Xn and a large function F The question is whether we can compute Such a protocol to compute f of The X1 up to Xn in time much less than the structure So that seems paradoxical at first because well you have to compute the function at some points But we actually managed to achieve something somewhat similar Where there's a first an offline preprocessing phase? so for the picture above Disconsisting compressing the function f and so everybody can do that offline. This is costly, but we don't care Then there's an online phase where people compute the ciphertext and that's our laconic Property suggests this is small. So the online complexity of the of the protocol here Of the part here is small and then people can individually decrypt the Decrypt the ciphertext. So in the end that this just gives an MPC where the online communication is small But that I think is pretty cool so you can More about that on the imprint or catch either hot take Daniel or me if you have any questions. Thanks for your attention