alphaspirit - Fotolia

News Stay informed about the latest enterprise technology news and product updates.

Bad news for encryption security, PKI certificate revocation

News roundup: Troubling research on PKI certificate revocation; encryption research finds usability lacking; GnuPG adds features. Plus: More zero-days, xCodeGhost still haunting Apple and more.

While one group of researchers reported that certificate revocation -- a fundamental feature of public key infrastructure security -- was almost completely dysfunctional, and just as open source encryption project GnuPG announced two new features to make its personal encryption program safer and easier to use, another research group reported that modern personal encryption software was virtually unusable by ordinary people.

And the one big cryptography success story this week was not good news: Dark Web vendors have been reported to be offering malware packages that include Stuxnet-style code signing certificates for signing malicious code, thus making it virtually undetectable.

Revoked certificates widely used, rarely verified

Researchers from Northeastern University in Boston, the University of Maryland, Duke University in Durham, N.C., and Stanford University in Stanford, Calif., reported that "a surprisingly large fraction (8%) of the certificates served" on live Internet Web servers have been revoked, which was bad enough, but they also found that most browsers don't check certificate revocation status, "including mobile browsers, which uniformly never check."

They wrote: "This uncertainty leads to a chicken-and-egg problem: Administrators argue that they need not revoke because clients rarely check, while clients argue they need not check because administrators rarely revoke."

Apple, Adobe and the U.S. government were found to be using revoked certificates; of the browsers tested for certificate verification, Internet Explorer 11 was the best at verifying certificates, while none of the mobile browsers correctly tested or validated TLS certificates.

The group also published research last year on the effects of the Heartbleed vulnerability, which it suggested was the reason many of the certificates found had been revoked.

The researchers found that certificate revocation lists (CRLs) "impose significant bandwidth and latency overhead on clients: The median certificate has a CRL of 51 KB, and some certificates have CRLs up to 76 MB in size."

One possible route to better distribution of certificate revocation is via the Online Certificate Status Protocol (OCSP), which allows clients to query the status of one digital certificate at a time, thus reducing the overhead of frequent CRL updates.

However, the researchers found that "OCSP Stapling, which addresses many of the difficulties of obtaining revocation information, is not widely deployed: Only 3% of certificates are served by hosts supporting OCSP Stapling." Even when OCSP is used, there is still a performance issue, as it "requires the client to delay accepting the connection until the OCSP responder can be contacted."

Another troubling result: They discovered 18 nonrevocable, top-level certificate authority (CA) certificates that can be used to issue trusted certificates for any domain. "Being unable to revoke a CA certificate is particularly worrisome, as possessing a CA certificate's private key allows one to generate certificates for any Internet domain -- and private keys for CA certificates have been inappropriately given out multiple times in the past."

They concluded: "Overall, our results paint a bleak picture of the ability to effectively revoke certificates today." However, they also note that there are ways to improve significantly in the near term, including by using an improved method for constructing certificate revocation lists to make them smaller, and thus, less of an issue for Web browsing performance.

GnuPG introduces a new trust model, DANE support for key distribution

Meanwhile, open source encryption project GnuPG introduced a new trust model, Trust on First Use (TOFU), which does what it says: It trusts that the key belongs to who says it belongs to the first time it encounters that key. The downside of this model is that there is no guarantee as far as the identity of the key holder; the only thing a recipient can trust is that the key is associated with the key's source. The new feature will ship in the next release of GnuPG and will be turned off by default.

Also newly implemented from GnuPG is support for DNS-Based Authentication of Named Entities (DANE), a protocol that uses "the DNSSEC infrastructure to store and sign keys and certificates that are used by TLS." According to GnuPG, "The basic idea is that users publish their keys in the Secure DNS. Then, when someone is looking up a key, they simply use DNS to find it."

DANE is implemented in the current release of GnuPG.

But even modern PGP is practically unusable for most users

Researchers at Brigham Young University in Provo, Utah, published Why Johnny Still, Still Can't Encrypt: Evaluating the Usability of a Modern PGP Client, in which they reported that "modern [Pretty Good Privacy] PGP tools are still unusable for the masses."

The 20 participants, grouped into 10 pairs, attempted to exchange encrypted email using the highly rated Mailvelope browser extension front end for PGP; only one pair -- the one with the single PGP-experienced user in the group -- was able to exchange encrypted messages in the time allotted to the task.

None of the participant teams were able to encrypt without making some egregious errors -- such as revealing secrets or encrypting with a public key rather than private key. And even the successful team needed the maximum time allotted, one hour, to complete the task -- and they also stumbled before succeeding.

Yet, not entirely bad for everyone

One place the public key infrastructure is working: the dark side. InfoArmor Inc., based in Scottsdale, Ariz., identified a new trend: underground malware vendors selling digital certificates for code signing. InfoArmor reported that the GovRAT malware includes the ability to take malware binaries and sign them with apparently valid digital certificates to defeat antivirus and other malware detection methods.

InfoArmor wrote, "Signed binary code is interpreted as potentially trusted and verified software, which allows it to slide under the radar of antiviruses and proactive defense systems." This technique is similar to that used for the Stuxnet worm and in the Sony attack, and previously had been associated with "state-sponsored cyberattacks."

InfoArmor reported that Dark Web vendors were offering authentic code signing certificates from CAs, including Thawte and Comodo for as little as $600. For the budget-conscious attacker, Dark Web vendors offered code signing of individual malware files for as little as $60.

In other news:

  • Google Project Zero spent a week working on zero-days for Samsung's Galaxy S6, and reported finding "11 high-impact security issues as a result." The Project Zero team members worked on three challenges: Gain remote access to contacts, photos and messages; gain access to user and location data from an application installed from the Google Play Store, with no permissions; and "persist code execution across a device wipe, using the access gained" from the first two challenges. Samsung has already patched eight of the serious vulnerabilities in its October Maintenance Release, and are expected to remediate the remaining three vulnerabilities in November.
  • Last month, Mozilla announced that it was considering speeding up its timetable for deprecating SHA-1 support; now, it's Microsoft's turn. Citing Mozilla's announcement, Microsoft said it, too, was now considering the possibility of an accelerated retirement timetable for deprecating SHA-1 support. Both companies are considering moving up the date after which their software will no longer accept SHA-1-signed TLS certificates to July 1, 2016, from Jan. 1, 2017.
  • Boo, again. XcodeGhost malware continues to haunt Apple. The malware is still active and has been updated to infect iOS 9. FireEye reported that the update, which it dubbed "XcodeGhost S," was active in 210 enterprises, of which 65% were in the education sector. Meanwhile, the CrytpoWall 4.0 update reported by Bleeping Computer encrypts file names, as well as the file contents, making it harder on victims to figure out what files have been affected.
  • To help pay for their merger with EMC, Dell is considering the sale of $10 billion worth of "noncore" assets, Reuters reported. Some of the Dell units that may be up for sale include security subsidiaries Quest Software and SonicWall; backup solution provider, AppAssure; and IT services provider Perot Systems.

Next Steps

Find out more about why PKI falls short of its goals

Learn how code signing works for Android and iOS mobile apps

Find out how certificate revocation lists are published in Windows Server 2008

Dig Deeper on PKI and digital certificates

Join the conversation

4 comments

Send me notifications when other members comment.

Please create a username to comment.

How does your organization manage PKI security?
Cancel
With stapling, then there is no wait for any OCSP server. What's more, OCSP servers have been placed on the fastest connections possible. I've found that when I forced OCSP (not right now), it was only Google's services that caused grief -- everything else worked perfectly well.
Cancel
Yes, OCSP stapling works, but the researchers reported that "Only 3% of certificates are served by hosts supporting OCSP Stapling."
Even if it's the most popular 3% of sites that use it, there are still a lot of other sites that don't.
Do you see OCSP stapling expanding from that 3%?
Cancel
Without apathy and with understanding of the problem, it sure can expand. The letsencrypt project might have some other benefits though, perhaps if that gets widely adopted, there may be scope to provide certificates that live for very short periods of time and are renewed constantly (maybe weekly, daily, hourly or some other interval that makes sense).
3% is a sad figure, but it also shows that there is plenty of room for improvement.
Cancel

-ADS BY GOOGLE

SearchCloudSecurity

SearchNetworking

SearchCIO

SearchEnterpriseDesktop

SearchCloudComputing

ComputerWeekly.com

Close