OpenSSL Security Advisory | Hacker News Reader

OpenSSL Security Advisory

mta.openssl.org

55 pointshannob22 comments

Open articleView on HN

> OpenSSL 1.0.2k users should upgrade to 1.0.2k

Well that's certainly helpful advice :)

Another OOB? At this point, hasn't anyone run a static analyzer on the darn thing and fixed all the warnings?

11/16 - NPR - https://mta.openssl.org/pipermail/openssl-announce/2016-Nove...

9/16 - dangling pointer - https://mta.openssl.org/pipermail/openssl-announce/2016-Sept...

5/16 - buffer underflow - https://mta.openssl.org/pipermail/openssl-announce/2016-May/...

3/16 - NPE, double free, etc - https://mta.openssl.org/pipermail/openssl-announce/2016-Marc...

... I'm tired looking...

Every time I've looked at the code, it's been … idiosyncratic to the point where I would imagine that “fixing all of the warnings” would be a significant exercise.

The OpenBSD team runs coverity on their LibreSSL fork and continues to improve the codebase constantly. More distributions are switching to LibreSSL too.

If static analyzers could find all problems in C with a low false positive rate, then that'd eliminate one of the big problems with C and partially negate the need for Rust.

I've run several respected static analyzers against several well respected projects and I'm lucky if I get more than 2-3 false positives.

In the Argon2 reference library, the one alert from Infer was a genuine bug I sent a PR for, followed by one alert from tis-interpreter that was another genuine bug. clang-analyser produces clean output.

It's true that you get so many false positives under OpenSSL it's not feasible to review. I would contend this is an indicator of code quality more than an indicator of poor static analysers.

I've written this article on CVE-2017-3730: https://guidovranken.wordpress.com/2017/01/26/cve-2017-3730-...

It includes proofs of concept and touches upon the fact that client vulnerabilities can affect some servers just as well, because servers sometimes initiate connections to other servers. Mail servers do this, as well as certain web applications.

For those looking around for near-drop-in replacement:

https://www.libressl.org/

Interesting to see the note about version 1.0.1 there. Isn't that the latest available in Ubuntu 14.04?

    Note
    ====

    Support for version 1.0.1 ended on 31st December 2016. Support for versions
    0.9.8 and 1.0.0 ended on 31st December 2015. Those versions are no longer receiving security updates.

I'm not intamately familiar how Ubuntu handles LTS releases, but I'd assume they don't backport security releases for such old third party packages? Or do they?

EDIT: I can answer my own question, they do backport security updates, even for unsupported software that they still ship

http://askubuntu.com/questions/863477/will-ubuntu-14-04-rele...

https://bugs.launchpad.net/ubuntu/+source/openssl/+bug/15966...

The most serious ones seem to be 32-bit only, which may provide you some relief / buy you some time if all your servers are 64-bit (though you should probably still upgrade).

I have OpenSSL update fatigue.

We all do. What distros is this affecting though? It seems like this is only in "recent" versions.

Aren't these the currently maintained versions? Anying pre-1.0.1 is no longer being maintained, and they note 1.0.1 will be going away soon too.

Quite right! I missed that bit.

    Note
    ====
    
    Support for version 1.0.1 ended on 31st December 2016. Support for versions
    0.9.8 and 1.0.0 ended on 31st December 2015. Those versions are no longer
    receiving security updates.

It's not clear whether the first CVE (crash via OOB read) affects 1.0.1.

The OpenSSL team may not be maintaining them but some distros are.

Sadly this may be a feature of all cryptographic software.

Isolate your exposed services, give them push button deploys to get current versions out to each node as quickly as possible.

> Isolate your exposed services

Most important part. vulnerabilities need to get patched, period. But I'm a lot less stressed about it when I know my publicly exposed surface is small, and even that has layers in front of it.

I don't think the code that Bernstein writes has many vulnerabilities.

what's the last version on jessy (debian) ? checking on my machiens says 1.0.1t . in one of the comment it seesm that 1.0.1 is not updated anymore. checking the apt-cache policy openssl shows installed and candidates as 1.0.1t-1+deb8u5.

What should I do?

This is a very useful site for Debian security issues:

https://security-tracker.debian.org/tracker/source-package/o...

so only unstable is secure.. what's the general approach then? Should I use sid? (this is controinutitive)

Reply on news.ycombinator.com