work

Most of this week was honestly pretty boring. There were a handful of admin meetings, and I also just had my time taken up with other research. There was nothing too exciting. I can already feel that I'm going to be pretty hard-pressed to finish both my (hopeful) STOC submissions, my FOCS presentation, and my Eurocrypt reviews. All of that is technically due in November, but I will have to handle it on top of my current class and other research projects. My goal is to stay on top of it these coming weeks, so that I don't get slammed close to the deadline.

The highlight of the week was definitely NY Cryptoday! It was the first time I'd been to the JP Morgan and Chase building, and it's always fun to see the views from the big Manhattan skyscrapers. (Even if the east-side of Manhattan is a rough commute from NJ :-).) I was really just happy to see the other cryptographers in the area and talk about what they're interested in and working on. I was especially glad to meet some current students at Columbia and NYU that are interested enough to continue working on cryptography. The talks were amazing too! I was most excited to learn about the new attacks against Classical McEliece, but I learned a lot about low-depth PRFs, zero-knowledge, and backdoors in classifiers too. It left me with a number of follow-up questions too.

non-work

I don't even know where the time went outside of work this week. I know I've been playing a bit more with GenAI. I made a pretty cool visualization of my acadmic geneology after my friend asked about it. I've also just been using it to do more literature search and understand things I'm curious about in and out of cryptography. I've also been spent some time working on fixing/cleaning Cryptology City, as well as building a local-only version that lets someone navigate the website from a local clone.

I also unfortunately went to inspect my furnace and ended up punching a hole through my ceiling that I need to get fixed soon. Fortuantely, it doesn't affect my new neighbors, who I also met a few time this past week. They're nice, and I think it'll be good in some ways to not be near a vacant apartment.

questions

• Why are quantum learners not useful for AC^0 or AC^0[p]?
  • There are some interesting classical barriers here, but it's not clear to me why there is less literature on quantum distinguishers
  • I think PRFs are generally just treated as unstructured enough to not admit quantum attacks, but at this low depth, is that still the case?
• Can one use the ideas for planting a statistically undetectable backdoor for building an open source watermark?
  • A backdoor would let you prove that the watermarked model was yours with a particular input.
  • But, could one extend this so that generative outputs can also be distinguished from non-backdoored models?
• What is preventing the McEliece attack for key distinguishing from provably extending to key recovery?
  • I think it's interesting that the authors rely on some conjectures for their real attack that would break CPA security
  • It doesn't matter if the scheme is broken in practice, but it would be nice to better understand the theory here.

Thanks for reading to the end! Here's a picture from the JPMC tower's 50th floor.