Thanks all for your feedback! I asked for mainstream use-cases for algorithms whose removal could cause widespread pain. Some individual users, undoubtedly, will be hit by this, and I acknowledge that they may not be reading this list. But I wanted to know if I'd missed something endemic. I also asked elsewhere: Adam Langley pointed me to the MD4 use-case and Steve confirmed that RC2 must stay. There is a tradeoff: by attempting to accommodate every single use-case, we will weaken the library for a substantial amount of our user base, by offering them bad defaults, or simply by virtue of the fact that our developer resources are not infinite. (Near)-dead code is a liability. So far, excluding suspicions that something may be used somewhere, I've received one clear argument, for RC5. And, while I'm sympathetic to the cause, I believe this is the case where we have to balance one user's needs against everyone else's. As for specific deprecation strategies: - Don't forget that all applications will have to recompile against 1.1. - Disabling algorithms doesn't work well for us as it's just pushing the decision on the distros. It also wouldn't win us any resources as we'd still be responsible for keeping the code bug-free. The only win would be in default compiled code size. - We can leave OPENSSL_NO_XXX defines around so wrapper libraries (Python, PHP etc) who correctly account for the fact that an implementation may be missing (which they have to, anyway) will continue to work. - Removing assembly support is a GREAT suggestion to simplify the complexity, and I think we should do this for anything we end up leaving in, and perhaps even for some things not yet on the hit list (RC4?). - I appreciate OpenSSL's role as a "Swiss army knife" research tool but research needs shouldn't prevail over those of real applications. We are generally not on the frontline of deprecating things - RC4 isn't yet on the list. SSLv3 isn't yet on the list, etc. But we can't become the Internet Archive of All Old Crypto either, and there's some cleanup to do that's long overdue. - I believe that the OpenSSL community generally tends to overlook the positive impact of change when trying to round up the negative impact of breakage. Applications are generally able to move along and fix things when presented with the right incentives. Applications that aren't generally able to move along are rather unlikely to jump onto OpenSSL 1.1, and all these algorithms will be supported as part of OpenSSL 1.0.2 until 2020 for them. Finally, removing support for these algorithms from OpenSSL doesn't render your encrypted storage inaccessible. You can always use another implementation to decrypt/re-encrypt your data, and I fully believe in the power of the community in providing such tools where necessary. - Recent anecdotal evidence: OpenSSL's Logjam protection caused pretty widespread MySQL breakage. That made MySQL backport a change increasing the DH key length from 512 to 2048 bits. For end user security, it's a win. I would prefer that we start cashing in these improvements before the next Logjam happens. (This is my view; as you can see views differ even within the team.) - On binary elliptic curves: with recent cryptographic advances, I believe these are now firmly planted in the "internet archive" category, even if not practically broken yet. The other reason for removing these is that our implementations are crappy. They're not constant-time, and they've been shown to contain bugs. Improving upon these implementations is not a good use of dev time imo, given their decreasing credibility. Here's the list of algorithms gone from BoringSSL: IDEA, MD2, MDC2, RC5, RIPEMD, SEED, Whirlpool, binary curves This isn't of course entirely representative of widespread usage. However Google's multi-billion line codebase now builds against BoringSSL and therefore largely does not depend on these algorithms. Those billions of lines aren't all new and shiny code written in 2015, and some of it does have to interoperate with the outside world. And here's the list gone from LibreSSL, from what I can tell: MD2, MDC2, RC5, SEED Neither have removed CAST, and there is presumably a good reason for that. (PGP?) It seems to me that these can pretty safely go: MD2 - (The argument that someone somewhere may want to keep verifying old MD2 signatures on self-signed certs doesn't seem like a compelling enough reason to me. It's been disabled by default since OpenSSL 1.0.0.) MDC2 SEED RC5 These could probably stay (C only): CAST IDEA RIPEMD (used in Bitcoin?) WHIRLPOOL as well as BLOWFISH MD4 RC2 I am on the fence about the binary curves: I am not aware of any usage, really, and it's not about to pick up now. Cheers, Emilia On Mon, Nov 16, 2015 at 2:21 PM, Hubert Kario <hkario at redhat.com> wrote: > On Friday 13 November 2015 14:40:33 Emilia K?sper wrote: > > Hi all, > > > > We are considering removing from OpenSSL 1.1 known broken or outdated > > cryptographic primitives. As you may know the forks have already done > > this but I'd like to seek careful feedback for OpenSSL first to > > ensure we won't be breaking any major applications. > > > > These algorithms are currently candidates for removal: > > > > CAST > > IDEA > > MDC2 > > MD2 [ already disabled by default ] > > RC5 [ already disabled by default ] > > RIPEMD > > SEED > > WHIRLPOOL > > ALL BINARY ELLIPTIC CURVES > > > > My preference would be to remove these algorithms completely (as in, > > delete the code). Disabled-by-default code will either be re-enabled > > by distros (if there's widespread need for it - in which case we > > might as well leave it in) or will be poorly tested and is likely to > > just silently rot and break. This code is bloat and maintentance > > burden for us - my hope is that much of this code is effectively dead > > and can be removed. > > > > *Are you aware of any mainstream need to continue supporting these > > algorithms in OpenSSL 1.1?* Note that an older OpenSSL library or > > binary, or a standalone implementation or another crypto toolkit can > > always be used to continue supporting a legacy standalone > > application, or to decrypt ciphertext from the distant past. I am > > looking for use cases that could cause e.g. interop breakage between > > new and old peers, or major pain to distro end-users. > > > > These algorithms are obsolete but removing them doesn't look feasible: > > > > BLOWFISH - probably still in use though I don't know where exactly? > > MD4 - used in NTLM > > RC2 - used in PKCS#12 > > > > *Did I miss anything from the list?* > > I'd say that this is the wrong approach. > > If you have old stuff signed with MD2, and then timestamped with MD5, > SHA-1, SHA-256 over the years, with new timestamp added every year this > makes the MD2 signature _valid_ and _secure_. > > If you have stuff that is encrypted, but in deep storage with any of > those algorithms, then yes, it's less than optimal. Removing support for > those algorithm will make this data inaccessible. Not to mention that > stuff like IDEA or SEED may never get broken before the data in them > needs to remain secret - and after that creating a new archive with > unencrypted data after it can become public would simply cost money. > > And as some pointed out, few OpenSSL users actually read mailing lists, > fewer still know what actual primitives are necessary for their stuff to > work (e.g. PKCS#8 files specify inside them the cipher and key > derivation necessary for decryption). > > > What I think is the correct course of action, is to have in the > configuration file settings which specify the set of algorithms that are > set as available and trusted. So that we can disable them by default and > require the users to make a concious decision to enable support for > them. (Make openssl print to stderr info about them being > administratively disabled when application tries using them for bonus > points). > > This allows to place the information about this depreciation in a place > where users will actually see it and will make them concious of using > weak and badly tested algorithms. At the same time it will protect most > of the users that don't require such obsolete algorithms. > > But this concious decision MUST NOT require recompilation of the > package. Few if any distributions support recompiled packages. For many > end-users this is also a hurdle they simply can't cross. > > And this also allows openssl to change the cryptographic policy in > stable branches without breaking the API/ABI promise. (POODLE, FREAK, > Logjam) > -- > Regards, > Hubert Kario > Senior Quality Engineer, QE BaseOS Security team > Web: www.cz.redhat.com > Red Hat Czech s.r.o., Purky?ova 99/71, 612 45, Brno, Czech Republic -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mta.openssl.org/pipermail/openssl-users/attachments/20151116/99af64b5/attachment-0001.html>
