Hacker News new | past | comments | ask | show | jobs | submit
xmcp123 | on: eMudhra to exit global SSL/TLS certificate business
The thing is I cannot even imagine a way that bug could exist that is not outright negligence.

A unit test becoming prod data? That’s really all I’ve got and that’s pretty unforgivable.

__MatrixMan__ | parent | next
It's easier to imagine if it was malicious.
jfyi | root | parent
It really is. The way I see it, you have two routes to go if you want a compromised key to use for your intelligence agency.

Do we pick something properly secure and nail our asses to the wall if it is ever found out, or do we pick something that looks like negligence so it can be shrugged off even if it makes discovery more likely?

Add to that eMudhra's quick exit, stage left. There is definitely something that smells rotten here.

kijin | parent | next
I've integrated with production APIs that tell the client to encrypt sensitive data using an IV consisting of 16 null bytes.

Some people just follow a spec and don't stop to think what that spec is supposed to accomplish.

shadowgovt | parent
Correct. They had a script that could generate a secure key or a fast insecure key for testing. The script generated both at the same time.

Human error was selecting the wrong output file and promoting it forward.

It's not at all the kind of thing you want to see in a reliable, secure, mature process... There shouldn't be a step where a human even makes a choice there.

And it's exactly the kind of thing most companies are doing, most of the time, more often than we would like.

An awful lot of the world runs on spit and bailing wire and things too easy to automate yet, and the parts that don't cost more.

cyberpunk | root | parent | next
Ever dealt with key generation on a HSM?

Thousand line shell scripts that use expect to drive 30+ answers to a multi-choice interactive tui (that only works on one version of redhat from 2008 or something) is not so unusual, can totally understand how it happened. No excuses in that though.

AlotOfReading | root | parent
Why would you need a fast insecure key at all though? Just use a real key. Key generation for an ECC curve should be lightning fast.