> Muhammad Usama Sardar <[email protected]> hat am 29.04.2026 
> 10:09 CEST geschrieben:
> 
> Also, I believe [non-hybrid ML-DSA] may be perfectly fine for constrained IoT 
> use cases 

Actually, I would make the opposite case. We have done measurements with Nordic 
devices for battery-constrained cellular IoT connecting via DTLS. These devices 
stay installed in place often upwards of 10 years. Both ECDHE and ECDSA 
authentication had a negligible impact on energy drain (in comparison to 
PSK-only). It turns out, both the power used for computation and sending the 
additional data for the ECC are tiny in comparison to the energy spent 
*listening* on replies from the cellular network. (If you can, optimize your 
send/receive patterns to match the heuristics of the network provider and 
modem. This is far more worthwhile than removing ECC from the handshake.)

Additionally, because updates are hard to deploy (see my previous mail [1]), 
IoT (and OT) has much less agility regarding authentication methods than the 
typical web examples (auto updated browsers and servers using ACME). In my 
opinion, IoT is therefore a prime use case for hybrid signatures.

> (and in such cases a good designer should actually use EDHOC rather than 
> TLS)

This is getting off-topic for this list, but I don't get why people like 
OSCORE. Mixing authenticated and unauthenticated data on the application layer 
is a giant footgun for the average software development team. What kind of 
infrastructure do people build that CoAP proxies are unavoidable? Isn't DTLS 
with SNI enough?

Best regards,
Tim Beckmann

[1] https://mailarchive.ietf.org/arch/msg/tls/pxcmlnlKe8u8sMrVk3pLErSlIBU/

_______________________________________________
TLS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to