Replace base64 with base64ct about cookie-rs HOT 5 CLOSED

Doesn't timing-attacks require physical access to the cpu that's parsing the data?
In that case isn't it a bit useless as you would practically have access to everything already?

I am probably missing something important though as it probably exists for a reason.

from cookie-rs.

Doesn't timing-attacks require physical access to the cpu that's parsing the data?

No. Timing attacks over the network (or any medium) are just as readily achieved.

from cookie-rs.

oh, had no idea that was practical, would have thought the differences around it (NIC buffering, routers, switches etc),
would make the range so big that it would be impractical decipher it.

It seems these techniques are more advanced than i thought.
In that case my take on this is to have it as an option (if the performance difference is "big").
And then give some practical example in the documentation on when the user should consider using the securer method/feature compared to the faster basic one:)

from cookie-rs.

Exploiting sidechannels as a network-based attacker is possible. It usually involves a fairly noisy attack which is repeated over and over many times (e.g. millions of times) combined with statistical methods to observe timing variabilities. Such attacks have been used in the past for full plaintext recovery against protocols like TLS e.g. Lucky13 (great paper if you want to see how such attacks work).

However, the most practical attack against a Base64 decoder/encoder is probably going to be a local microarchitectural sidechannel. Such sidechannels have been used in a research setting to recover Base64-encoded cryptographic keys and could still be possible against e.g. a local webserver (possibly on a cotenant container or VM): https://arxiv.org/pdf/2108.04600.pdf

from cookie-rs.

Closing with my commentary at #196 (comment).

from cookie-rs.