OpenSSL Receives FIPS 140-2 Validation
Argon writes "Close on heals of NewsForge reporting about Government Agency dragging its heels on OpenSSL validation comes the news that OpenSSL receives FIPS Certification. More details are available at the Open Source Institute site which has been driving the effort to get OpenSSL certified.
FIPS 140-2 certification allows software using the certified version of OpenSSL to get into various Government departments previously not possible, thus increasing penetration of Free Software in Government."
Congrats and thanks to the team - I can only imagine what a struggle this has been.
o licy_FINAL.pdf
p s-nist-update_oss-institute.org
From http://www.oss-institute.org/
Two points to remember please: a) the validation is still considered
"pending" until it is posted on the NIST site...in no more than 2
weeks from the announcement date -- NIST official protocol, and b)
the validation does not immediately solve all FIPS 140-2 compliance
issues.
The big thing available now is "OpenSSL Security Policy Version 1.0"
http://oss-institute.org/images/OpenSSL_SecurityP
This document is required as a part of the FIPS 140-2 validation
process. It describes the OpenSSL FIPS cryptographic module in
relation to FIPS 140-2 requirements. The companion document
OpenSSL FIPS 140-2 User Guide (Reference 14)is a technical
reference for developers using, and system administrators
installing, the OpenSSL FIPS software, for use in risk assessment
reviews by security auditors, and as a summary and overview for
program managers.
The "validated OpenSSL USER GUIDE" will be available within two weeks
of the announcement date.
No sign yet of OpenSSL 0.9.7j on the openssl site.
There is an email list available for updates:
http://mail.oss-institute.org/mailman/listinfo/fi
--Neal
Go IETF!
Could you cite your sources? From what I can tell, the FIPS 140-2 list of Approved Security Functions includes AES, and Triple-DES, as well as (curiously) DES and Skipjack[1].
For AES, the ciphers can be operated in the ECB, CBC, CFB, OFB, CTR, CMAC, and CCM modes of operation.
Approved hash functions include SHA-1, SHA-224, SHA-256, SHA-384 and SHA-512. Keyed hashing must be done using HMAC, but you can use various DES MACs, as well as CCM mode, for message authentication.
Interestingly, what this basically means is that FIPS 140-2 compliance does not imply that your system is secure. All it means is that the government can use it.
[1] Can somebody please check this? I vaguely remember DES and Skipjack being withdrawn, but I can't find the documentation for that.
http://outcampaign.org/
As a side note, it never seemed as if Microsoft's failure to get CC validations promptly ever slowed down IIS or XP deployments, but it's been a major roadblock for any other systems to get through DITSCAP if there was any possible reason to deny the request.
FIPS accreditation removes the final roadblock for open source in the federal government. Now there is not a single valid policy or security requirement that can block deployments of open source systems.
Also of note is that since anyone can use OpenSSL, small development shops are no longer held hostage to Certicom's expensive licensing schemes if they want to deploy FIPS compliant solutions. It used to be financially daunting to sell software to the government that included crypto, and this created a nice, safe sandbox for the small set of approved vendors to charge outrageous prices for FIPS compliant solutions. Now they have to compete with open source, which will likely bring costs down considerably for anyone required to deploy only FIPS compliant solitions.
Another poster mentioned that this restricted the choice of encryption algorithms to 3DES. That is incorrect. FIPS 140-2 is an AES implementation, specifically because of concerns over 3DES' long-term viability. There are no approved 3DES implementations under FIPS 140-2.
OpenSSL is one of those cool projects that would be so much cooler if it weren't for the stupid license that makes it a PITA to actually employ in a product. OpenSSL essentially uses the BSD license w/attribution, which makes it difficult to use with GPLd projects, unless you use the version provided by your distro -- which isn't always desireable.
Okay, maybe this is a question of semantics, but since when did a non-viral open source license qualify as "annoying"?