crypto

TEEs are the best technology today for speed, privacy, and decentralization.

They are particularly salient for L2s since L2s optimize for different things than L1 and as a result are harder to decentralize in the typical ways. In particular, super fast UX or novel ordering schemes like prio fee ordering.

Working on Farcaster and Warpcast. Longer thoughts at https://dwr.email

Aren’t there massive security concerns surrounding TEEs? This seems like adding additional central point of failure risk to an already hyper-centralized solution 🤔

Perhaps I misunderstand what their use case though.

Rated 💯 by multiple networks. Fluent in /unglish. Unbothered by /whatday of the week it is.

Allegedly a designer.

the problem is that this is a software guarantee, if the TEE is addressable by any means, extraction of key material is a matter of how, not if.

These are all separate SGX breaks, from four different years:
https://www.intel.com/content/www/us/en/security-center/announcement/intel-security-announcement-2024-08-29-001.html
https://cubist.dev/blog/intel-sgx-is-broken-again-what-the-downfall-attack-means-for-secure-hardware
https://arstechnica.com/information-technology/2022/08/architectural-bug-in-some-intel-cpus-is-more-bad-news-for-sgx-users/
https://sgaxe.com/files/SGAxe.pdf

They make sense for defense in depth where the trust boundary isn't being extended by the TEE itself, but if they _become_ the trust boundary, bad things happen.

BDFL of Quilibrium, Eng @ Farcaster, ex-Coinbase, always opinionated, never hydrated

/quilibrium

Appreciate that context. Frankly, it seems like a matter of time before this becomes problematic. Between this and the obvious risk that most of these L2s are secured by multisigs with a few common owners… I’m not confident. At some point the trade offs become too risky, no?

They make sense for defense in depth where the trust boundary isn't being extended by the TEE itself, but if they _become_ the trust boundary, bad things happen.

https://warpcast.com/cassie/0x085b36f1