[Cryptography] Complexity vs. security

[Peter Gutmann]

This recently came up on a (public) mailing list but I thought it's relevant
here too.  Quoted from djb:

>https://arxiv.org/abs/2107.04940 studied hundreds of CVEs in crypto
>libraries. There was roughly 1 new CVE per 1000 lines of new code. About 1
>out of 8 "cryptographic" CVEs was severe.

An interesting quote from that study:

  we find evidence of a strong correlation between the complexity of these
  libraries and their (in)security, empirically demonstrating the potential
  risks of bloated cryptographic codebases

I thought this was very apropos considering the neverending stream of codebase
bloat enhancement drafts posted to said mailing list.  One of the major skills
needed with security/crypto standards is knowing which parts of them, and in
some cases which entire standards, you can and in fact need to for security
purposes, ignore.  I don't mean the "check that the point is on the curve"
stuff but the "here's a cool idea I had in the shower this morning" stuff.

This is why security developers are at least one thing that AI will never put
out of a job, you need actual skill to sort the mess out, not just an ability
to extrude code from a spec.

If working groups like TLS and LAMPS adopted the rule "every time you add
something you need to take something else out" we'd have vastly less
complexity/bloat-driven vulnerabilities.  And, with the elimination of the
red-queen problem, stable standards where you can spend the time to get the
implementation right knowing it won't have some new piece of complexity bolted
onto it next week, and the week after, and the week after that.

Peter.