Are there any significant ways in which TLS could still be improved?

Collin

12/23/23, 8:05 PM

With TLS 1.3 supporting only secure, forward-secret cipher suites, are there any significant technological improvements that could still be made to the protocol?

1 + 1

block-cipher

forward-secrecy

poncho

12/23/23, 8:43 PM

The problem that TLS attempts to solve has many components to it, some conflicting at times - the designers of TLS have selected one particular set of trade-offs, which (depending on the criteria you think is important) may not be optimal. Could someone select a "better" set of trade-offs? Most certainly (especially if you give different weights to the various goals).

Score:1

Crypto

high_byte

12/23/23, 8:44 PM

In terms of technological improvements, yes.

Putting cryptography strength aside - I'm working on zero-knowledge proofs for HTTPS oracles and the biggest issue is TLS uses AES, a symmetric encryption. This makes it impossible to distinct between server responses and client request, forcing to make trust assumptions that the prover is checking the correct side of the conversation. (since the client & server share the AES key, the client could send a packet that looks exactly one the server could send)

There are some less-than-perfect workarounds to this, but in terms of technological improvements this one is easy to implement and has great practical implications, but unfortunately is probably going to take a long time for wide-spread adoption.

To summarize: Adding data signing to TLS session data is one such improvement.

+ 4

poncho

12/23/23, 8:54 PM

"the client could send a packet that looks exactly one the server could send"... "Adding asymmetric encryption to TLS session data"; actually, using public key *encryption*, both sides can still generate the same encrypted traffic - perhaps you meant public key signatures?

high_byte

12/23/23, 9:01 PM

yes that is correct. thanks

Marc Ilunga

12/23/23, 9:40 PM

Can you expand on "HTTPS oracles" ? In any case, I have a hard time seeing the computational overhead would be widely accepted and if such a functionality can justifiably be baked into such fundamental components ( authenticated key exchange and secure channel protocol)

high_byte

12/24/23, 2:44 PM

the first use-case that immediately comes to mind is a general purpose web2<->web3 bridge. imagine querying a JSON API and having the response on the blockchain with security equivalent to the web2 (HTTPS) server. e.g. getting a tweet, weather report, database query.

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: Are there any significant ways in which TLS could still be improved?

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.