Forgot your password?
typodupeerror
Security IT

Do the SSL Watchmen Watch Themselves? 171

Posted by Soulskill
from the barbers-with-a-good-haircut dept.
StrongestLink writes "In an intriguing twist on the recent Comodo CA vulnerability discussed here last week, security researcher Mike Zusman today revealed that three days prior to StartCom's disclosure of a flaw in a Comodo reseller's registration process, he discovered and disclosed an authentication bypass flaw to StartCom in their own registration process that allowed an attacker to submit an authorized request for any domain. During a month which was marked by the continuing paradigm shift to SSL-verified holiday shopping, the Chain of Trust continues to run off the gears, and Bruce Schneier is even commenting publicly that SSL's site validation mission isn't even relevant. What lies ahead for the billion-dollar CA industry?"
This discussion has been archived. No new comments can be posted.

Do the SSL Watchmen Watch Themselves?

Comments Filter:
  • by coryking (104614) * on Friday January 02, 2009 @10:05PM (#26307853) Homepage Journal

    SSL certificates are one area best served by government. Bear with me here,

    SSL certificates are the online version of your driver's license or your passport. We entrust our governments to provide us with reliable, trustworthy forms of identification. We know that if we see a driver's license or a passport, we can be reasonably certain the person holding said identification is who they claim.

    It is becoming increasingly clear that SSL certificates issued by private industry cannot be trusted. Since private industry issues them, there are real standards for how one qualifies for a certificate. A $20 SSL cert from Godaddy is just as valid of identification as a $500 one from Verisign. Worse, the private industry has a conflict of interest. Their business makes money by issuing certificates to paying customers, not rejecting customers for bad information. The more stringent their policy, the more applicants they reject, and the less money they make. It is simple math, they have to make it as easy to get an SSL certificate as possible or go under. (The bond rating industry suffers from a different, but somewhat similar conflict of interest, actually)

    Who then should issue certificates? The only entity that doesn't have to make money--your governments. Ideally you should be able to walk into whatever agency issues photo identification in your country and somehow get an SSL certificate issued. Businesses and non-profits could get them issued by checking a box on the form they use to set up a corporation or LLC.

    Letting the government deal with this has many extra benefits. For starters, we could make SSL certificates fall under the same kinds of laws that govern passports or drivers licenses. If you forge one, or enter fake information, you could be charged under the same laws that faking a drivers license fall under. For second, if done right, good governments would issue these for virtually nothing and maybe protocols like S/MIME would finally get widespread adoption.

    What about open source projects who currently cannot afford SSL certs? Well, if the government does it, they could file as a non-profit and get one for free (or reduced cost).

    How would this work from a technical standpoint? How would browsers deal with a long list that has every countries certificate authority? Dunno, but it seems it wouldn't be a big problem. It is a technical problem though, so we can solve it somehow.

    What international agency would regulate this? Who regulates passports? Dunno, but seems to me we already have a long history of internationally recognized identification--both for business and personal use. Why not task those guys with SSL certificates? This is more of a political problem, and isn't as easy to solve as the technical bits.

    Bottom line, I know we all seem to hate more government, but SSL certificates are one thing governments should be doing, not private industries. It might create a new class of problems, but I suspect the new problems will be much less severe than the ones we have now.

    • by Anonymous Coward on Friday January 02, 2009 @10:12PM (#26307915)

      I can't wait to see the phishing websites validated by the Nigerian government's CA.

    • by wizardforce (1005805) on Friday January 02, 2009 @10:13PM (#26307927) Journal

      Your trust of government is simply astonishing after what the Bush administration has been up to for the last eight years especially considering all those slashdot stories concerning fumbling incompetence on the part of certain governments... The problem wish computer security isn't private industry, it's that there are few direct consequences for companies that produce faulty security systems, banks with shoddy security etc.- legally granted limited liability is a problem, Once they find their own heads on the chopping block after a security flaw is found they'd be a lot more keen on solving the problem.

      • by Cowmonaut (989226)

        Hey now, don't belittle the strengths of a bureaucracy because of Bush. There are certain things it can do well, licensing is one of them. It's not perfect (not hard to get a fake ID) but its good enough (moderately difficult to get a GOOD fake ID). Plus, then you know for sure that someone is checking on the security of the certificates because that's 50% of their job.

        Now if only they'll make it so where there is a road, there is pipe (for the most part) and get some of the boonie yahoos some decent DSL

        • Plus, then you know for sure that someone is checking on the security of the certificates because that's 50% of their job.

          Yes, and there are also supposed to be people making laws that agree with the constitution and striking down unconstitutional ones, and people that make sure patents are valid before they get approved. But in both of them they fail in their jobs.

          And think about the ways that governments would abuse this system. For example AT&T might not have a decently secured site, but because they agreed to wiretap they might give them a certificate. On the other hand a site that sells materials disagreeing wit

        • Plus, then you know for sure that someone is checking on the security of the certificates because that's 50% of their job.

          Don't be so sure:
          http://www.computerworld.com.au/index.php/id;50110485 [computerworld.com.au] they [at least the UK] seem to be fairly adept at losing things, if they screw up big time you still pay for it.. when a company screws up bad enough at least people might have a chance to look elsewhere- no, I think the solution here is to make use of that horrible trait of human nature- greed, well at least enli

          • Exactly. As an end-user (businesses refer to you as a consumer), you expect
            that the website you are interacting with is who you *trust* them to be.
            And as the end-user, you expect that the reason you trust the site is because
            you have the lock showing in your browser, and you believe the SSL system
            is trustable.

            Yet, as the end-user, what have you personally seen as evidence
            that the https protocol using SSL is really trustable?

            Most people have seen nothing.

            And yet, here someone says the government should be tr

            • So you trust your government less than a random company that has bought its CA status with money?

              • given enough competition I trust that if any of them prove themselves unworthy of trust that it's still a better system than any of our governments could design and no I really don't trust a governmental monopoly over competitive private industry especially when our little government has been caught spying on its own people.

                • Would you care to elaborate on how a private company is supposed to compete for trust and profit at the same time, without sacrifying one for the other?

                  Oh and btw: A governmental CA can not be used to "spy" on anyone. Put down the tin foil...

                  • by JoelKatz (46478)

                    "Would you care to elaborate on how a private company is supposed to compete for trust and profit at the same time, without sacrifying one for the other?"

                    You can't sacrifice trust for profit if you're in the trust business. All you have to sell is the trust people have in you. Give that away and you have nothing to sell.

                    • I take it you have never order a SSL certificate at a shop like RapidSSL, instantSSL and the ilk?
                      If your definition of trust translates to "owns (or stole) a credit card" then yes, today's PKI is perfectly fine.

                    • by JoelKatz (46478)

                      All I'm trusting them to do is verify the domain. They do that.

                      Making it harder to get any certificate at all just means that less traffic will be encrypted. That's a much worse problem than man-in-the-middle attacks.

                  • why on earth would you continue to use an untrustworthy company's product if there is any competition whatsoever? why would a company stay in business making profit if people abandon them for more trustworthy companies? If a company wants to make as much money as possible [profit] and being untrustworthy undermines that profit base, why would they be able to continue? the only situation where things could not improve in this manner is with a monopoly, private or public monoplies have no incentive to impr

              • by Znork (31774)

                As basically every government is, or wants to, listen in on any traffic they can I don't only not trust them, I am utterly certain that they will issue any number of falsified certificates enabling them to intercept and MITM any SSL communication they want to. The CA's have yet to indicate that desire. Not that I think most would hesitate to sign a false certificate on request from the government anyway.

                So for the purpose of certificates, I trust governments far less than a random company. Of course I also

                • I am utterly certain that they will issue any number of falsified certificates enabling them to intercept and MITM any SSL communication they want to.

                  Well, if that is your concern then you are just utterly clueless about how SSL works.
                  The CA can not spy on your SSL traffic, no matter how much they want it.

                  • by AnyoneEB (574727)
                    The CA cannot spy on any arbitrary SSL traffic, but they can MITM any SSL connection (that they can intercept) because they are able to create a trusted certificate that says anything they want so they can make their own key pair and sign that. Then they can spy on any connection they MITM. More work than just looking and theoretically detectable, but CAs certainly have the ability to spy on SSL connections if they want to.
                  • by Znork (31774)

                    You need to look up how man in the middle attacks work. As long as they can create a signed certificate saying their server is the destination server they can transparently proxy your communications.

                    For non-government CA's it's tricky, as they'd have to spoof or control DNS for the domain in question, but for a government it would be trivial. There's already many examples of redirects and censorship proxying being done, and as long as a government can either produce, or compel to be produced, signed certifi

                    • Claiming that such a scenario would be "trivial" for a government or anyone else is just nonsense.
                      The government does not own DNS, nor does it own the ISP pipes. If they want to go to such lengths as to fake a CA they can just as well rubberhose a privately owned CA into doing it for them today - and it would probably be much cheaper on a per-case basis than permanently maintaining the required infrastructure themselves.

                      Furthermore, what interest does the government have in snooping on our online banking an

                    • by Znork (31774)

                      Claiming that such a scenario would be "trivial" for a government or anyone else is just nonsense.

                      Many governments already do siphon off traffic to proxies and/or interfere in ISP DNS services for child porn blocking, as well as for various intelligence purposes. To think that's hard to do when it's already done is rather disingenious.

                      they can just as well rubberhose a privately owned CA into doing it for them today

                      They probably do it regularly. I can't see Verisign objecting to a national security letter.

    • by djupedal (584558) on Friday January 02, 2009 @10:14PM (#26307941)
      >SSL certificates are one thing governments should be doing

      So, after wading patiently thru your treatise, it would seem you elected not to answer the question, which would explain your warmth towards politicos, at least :)
    • by Cyberax (705495) on Friday January 02, 2009 @10:15PM (#26307945)

      It's better to use private companies with government oversight.

      I now live in Ukraine and we have such a system. Government licenses private companies to work as certification centers and mandates that only certain (strong) crypto algorithms must be used.

      As a result, I can use my private key to sign my tax report for IRS (or tax report for my company). IRS in turn uses its own key to sign their letters.

      That's pretty cool, if you think about it.

      • by iammani (1392285)
        Its the same in India, I can file my taxes online, and sign them with my private keys (Issued by 3 authorized private cos) or print the confirmation, sign and hand it to the tax office physically.
    • Re: (Score:2, Insightful)

      by minsk (805035)

      So you have some governments that issue high-quality reliable certificates.
      And some corrupt ones which can be bought for peanuts.

      So someone has to choose which root certificates to trust.
      Someone, probably being the browser makers.

      So what would it solve?

      • You'd have the browser show which country issued the certificate. Use a flag, use something. Firefox already does this by using a tooltip.

        Plus, unlike private companies, we all have a sense of which countries certificates we may or may not trust. A user would get suspicious if "bofa.com" was using a certificate issued by Nigeria or "tesco.com" had a certificate that wasn't issued in the UK. What the fuck is the difference between a certificate issued by Thwarte vs. Verisign? Beats me!

        • You're putting a bit too much faith in the user I think.

          Perhaps if the browser stored every certificate the first time it was seen, then flagged the user when it was changed (combined with relying on certificate chains and the like) we wouldn't be having so many issues with MiTM.
          • No and yes. I think both of you are making great points.

            The flags are a great idea because they give the users who care a meaningful tool to assess the trustworthyness of the site at hand.
            Knowing the country of origin is much more meaningful than an anonymous padlock.

            Saving the cert fingerprint and raising an alarm on change is not even a great idea by any means - it is just obvious, absolute baseline stuff.
            The Mozilla guys are seriously humiliating themselves by fucking up the SSL handling even more instea

      • by Cyberax (705495)

        So we need some way to rate CA quality...

        Also, we can consider using money to fix this problem. For example, we can make all CAs put a big sum of money into an escrow account to be given to the first person who shows that CA doesn't perform 'due diligence' while issuing certificates.

    • by Phroggy (441) <slashdot3&phroggy,com> on Friday January 02, 2009 @10:39PM (#26308107) Homepage

      It is becoming increasingly clear that SSL certificates issued by private industry cannot be trusted... Who then should issue certificates? The only entity that doesn't have to make money--your governments.

      The problem with your idea is, even though you're correct that private industry cannot be trusted in this matter, the government cannot be trusted in this matter either.

      These are technical flaws, not policy flaws - mistakes are happening due to software errors, NOT because some executive decided that allowing anyone to have a certificate without verification would be a great idea. I may trust the government's intentions, but experience suggests that they won't develop a system like this in-house, but contract it out to the lowest bidder, who is likely to have far less experience with this sort of thing than the current players.

      For starters, we could make SSL certificates fall under the same kinds of laws that govern passports or drivers licenses. If you forge one, or enter fake information, you could be charged under the same laws that faking a drivers license fall under.

      Pretty much all current spam is illegal under the CAN-SPAM act, so spammers could be charged under that law. They're not. I have no confidence that fake SSL certs would be prosecuted.

      • I have no confidence that fake SSL certs would be prosecuted.

        Do governments crack down on people who fake their passports? If so, what is their motivation for doing so? How would their motivation for cracking down on SSL forgeries be any different?

        • by Fjandr (66656)

          How would their motivation for cracking down on SSL forgeries be any different?

          You can't transport someone into a country with a fake SSL cert.

    • by mortonda (5175)

      Who then should issue certificates? The only entity that doesn't have to make money--your governments.

      Specifically, I would opt for Notary Public, maybe as a specially trained office, but the function is nearly identical.

    • I couldn't be bothered to read this whole thing at 3am but I will say this. There is no reason a $20 cert from GoDaddy is any less valid than a $500 verisign one. The largest difference is one is making you pay extra just as you would for a sports car but in the end both get the job done.

      Lastly, trusting the government not to cock this up relies on all countries doing the same thing and it relies on governments sorting their acts out and stop fucking things up as virtually every government seems to do.
    • What about simply creating a better web of trust? For example, if you only trust governments, then you only accept certificates issued by them. If I trust Verisign but not Godaddy, then I only accept Verisign and the other sites I trust.

      This is how a web of trust should work. People trust certain sites to issue certificates. As certain sites gain trust, more people want to get certificates from them, etc. I might trust my friend Bob, but there is no reason you should. If a bank or e-commerce site wants t
      • What about simply creating a better web of trust?

        Congratulations!

        You Sir have just re-invented CaCert [cacert.org]. CaCert is a certification authority which operates by a web-of-trust model: users certify each other after seeing id, and only users having gathered a minimum amount of assurance points can get a certificate.

        Unfortunately, CaCert is not trusted by the browsers (such as Mozilla or Konqueror), who seem to be more hung up about expensive audits and pompous root key signing ceremonies.

        Other CA's, such as Comodo/CertStar [mozilla.org] or RapidSSL/GeoTrust [win.tue.nl] don't seem to

    • by Lumenary7204 (706407) on Friday January 02, 2009 @11:39PM (#26308525)
      The United States under the Clinton/Gore administration already tried something similar to this; five words spring to mind: "Clipper, Skipjack, and Key Escrow". (If you need a refresher, I suggest the book "Crypto" by Steven Levy [amazon.com].)

      The **last** thing I want is for my government to be the entity that issues the requisite public/private key pairs to the private institutions and companies with whom I do business. My business is **my** business - and not the government's business - until a **legitimate** search warrant or indictment says otherwise. And even then, it's still **my** business [wikipedia.org].

      As the article posting indicates, SSL is built around a Chain of Trust. People buy SSL certificates from the likes of VeriSign, Thawte, Equifax, etc., because they are well-known and (ostensibly) trustworthy organizations.

      I, for one, do not entirely trust my government. I don't trust VeriSign and crew all that much, either, but their reputations are a strong motivation for them to do their jobs reasonably well, and provide products that perform as advertised. To do otherwise would damage their reputations, resulting in lost customers and weaker profit margins.

      Most governments, on the other hand, don't care much about their reputations, and have little regard for profit margins (just look at the US Government's annual budget deficit). They therefore have no compunction against using excuses such as "national security" and "protect the children" to provide (at best) or mandate (at worst) inferior solutions to technological problems.

      Admittedly, some companies - like AT&T [wikipedia.org], for instance - are so large and well-entrenched that they sometimes bow to the mandates of government, and little heed the damage done to their reputations because of it.

      But most companies are not that large, and can ill afford to lose face in the marketplace. Reputation is their bread-and-butter, so they do what's in their own best interests, which may even coincide with their customers' best interests.
      • by sjames (1099)

        Clipper etc. was a scheme where a back door was explicitly built in. A system where the government signs your PUBLIC key without ever seeing your private key wouldn't permit such abuses.

        That is part of the solution. In addition, the web of trust needs to be more configurable in any case. I may trust a particular key's validity. I might or might not trust keys signed by it. Further, I might trust that much but not trust keys signed by a particular key to sign other keys (I know the key belongs to the person

        • by Znork (31774) on Saturday January 03, 2009 @10:33AM (#26311423)

          without ever seeing your private key

          Why would they need your private key? As long as they can sign any key as being valid for being 'you' they can make their own signed public/private key pair purporting to be you and MITM any communications to you. To get around that you'd still need out-of-band exchanges of the keys in which case the government signing serves no purpose.

          In addition, the web of trust needs to be more configurable in any case.

          Without a doubt.

          • by sjames (1099)

            Why would they need your private key? As long as they can sign any key as being valid for being 'you' they can make their own signed public/private key pair purporting to be you and MITM any communications to you. To get around that you'd still need out-of-band exchanges of the keys in which case the government signing serves no purpose.

            They would only need my private key if they wanted to implement the 'key escrow'/clipper sort of snooping. My only point was that just having the government sign your public key doesn't enable that.

            I do agree that there are cases where a government might impersonate someone itself. That's where the configurable web of trust comes in. You might even want to add the concept of scenario and assign different levels of trust based on the scenario. For example, if I'm browsing the website of a Chinese business I

      • by Eskarel (565631)

        That's not really what signed certs are for though.

        You don't really use your signed cert to encrypt your data(for data encryption you don't need a signed cert, and additional information is used within the SSL procedure to generate temporary keys. I can get a copy of the signed cert for your bank, but that doesn't mean I can read the transaction you're making. You don't even have to have a signed certificate to have secure transmission of data.

        Signed certs are about validating "who" someone is, they are pre

      • I agree with sjames when he disputes your comments regarding similarities between Clipper and X.509. Except that they both concern cryptography, there are no similarities between them.

        The Clipper fiasco was a failed attempt by US government to build a deliberate back door into a specialized crypto algorithm. It was doomed for many reasons, not least because governments such as mine pointed out that it would not be in our national interest to import products which used the Clipper chip. Realizing that
      • by ckedge (192996)

        Do you know who I'd trust with this kinda thing? Certain groups of people in the NSA*, NASA, and Education Instututions. If anyone knows how to make something bulletproof, if anyone knows how to do category-5 (CMM) software development, it'd be people from the first two agencies, and people from the latter are both smart and already incented to do things for the good of mankind. And it shouldn't be US centric, of course, everyone in the world should sign on and provide some money and people (I'm a Canuck

    • Which Government? (Score:2, Insightful)

      by upuv (1201447)

      You have placed your trust in the government. However which one?

      Most governments would with the best of intentions try to do the right thing. However some would not. Some would down right look at this as a cash cow. It would be ripe for the picking of corruption and miss use. With next to no legal recourse.

      So who governs the government?

      I would contend that this belongs in the hands of grander body. The UN or blocks of countries, the EU, NAFTA, African Union, G8,9,10,11(What ever it is now). etc. At le

      • Ideally, more than one the UN+ a local country competing would be better than either on their own.

      • These are valid questions, no doubt. Who oversees passports? I'd look real closely at how those get handled and steal the bits that work for them. There is a lot of overlap between the two.

        With next to no legal recourse.

        Once governments handle SSL, this becomes politics on an international level just like trade. If those damn Canadians don't stop with the crappy certificates, we Americans will just stop buying their maple syrup. Or something like that.

        That said, ultimately "legal recourse" always disti

    • by Plutonite (999141)

      The more stringent their policy, the more applicants they reject, and the less money they make. It is simple math, they have to make it as easy to get an SSL certificate as possible or go under. (The bond rating industry suffers from a different, but somewhat similar conflict of interest, actually)

      It's never that simple, clearly, because there is another factor called "trust". If you let in too many false positives, you lose the trust hierarchy and are pushed out of business by the other (more stringent) competitors. Who will put the government out of business when their sloppiness leads to disasters(as it uniformly has when dealing with security)? We trust the government locally because federal/state docs are produced with other federal/state documentation - we have 'faith' in the authentication mec

    • >> We know that if we see a driver's license or a passport, we can be reasonably certain the person holding said identification is who they claim.
      >> but seems to me we already have a long history of internationally recognized identification--both for business and personal use.

      Apparently no. That's the reason the travel to USA is now a PITA with all that added biometric registrations.

      And for developing countries, the passports never were enough: because immigration laws, most require visa applica

      • My nightmare is a bunch of companies with massive conflicts of interest issuing bullshit certificates. Nobody but nerds understand SSL and Mom and really even myself cannot tell what makes a good certificate "good".

        I also think government SSL would actually increase innovation in other, more productive industries. Government issued SSL certificates would most likely mean everybody gets ones. That means things like S/MIME become widespread and SPAM gets harder. That means code signing becomes widespread

        • Governments never agreed in a single worldwide way to generate trivial sequential numbers (for example, SSN are useless outside USA) and I find a bit impractical (from a political POV) that they can agree on a single scheme for something a lot more complex (and potentially dangerous) like SSL certs.

          And in that hypothetical scenario, be ready for USA banning Cuba CA's (and all current enemies); same the other way against USA; and also more bans against several categories of immigrants (as currently happens w

    • Re: (Score:3, Interesting)

      by pha3r0 (1210530)

      Their business makes money by issuing certificates to paying customers, not rejecting customers for bad information. The more stringent their policy, the more applicants they reject, and the less money they make. It is simple math.....
      Who then should issue certificates? The only entity that doesn't have to make money--your governments.

      Sir. I am not sure where you live but here in America we have seen countless changes made by various government agencies just so they can grab more tax money for there already inflated budgets.

      Allow me to weave a tale for my fellow readers. My very first job was in a paper and printing supply warehouse. Things were great. I worked there for about 6 months before I got a rather strange call. It was a customer of ours who placed regular orders for pens and toner and the like. She said she was going to be plac

    • Your overall point is rather silly, but this in particular stuck out:

      Worse, the private industry has a conflict of interest. Their business makes money by issuing certificates to paying customers, not rejecting customers for bad information. The more stringent their policy, the more applicants they reject, and the less money they make. It is simple math, they have to make it as easy to get an SSL certificate as possible or go under. (The bond rating industry suffers from a different, but somewhat similar co

      • Let me clarify my last statement:

        You make it, there is no guarantee that someone won't end up breaking it, or find some flaw or way around the system.

    • by DarkOx (621550)

      You are totally wrong giving the problem to governments does nothing to address the trust issue. Is a cert from Libia as good as one from the UK? How could the average person know other then by using applying the same international prejudices we use today for other things? How is that any different then trusting Godaddy more or less then Verisign?

      The problem is a certain popular web browser shipping with windows and the most popular open source browser for following the behavior of the former ilk. They

    • by wkk2 (808881)
      I'm sure governments would also like to generate your private key while they issue certs. Maybe a middle ground would be to have government enforced standards with audits on the CAs.
    • by JoelKatz (46478)

      "The only entity that doesn't have to make money--your governments." It's also the entity that suffers the least for its mistakes.

    • Be careful what you wish for [luxtrust.lu].

      The result:

      • Usage of this CA will be compulsory for securing interacting with the government
      • Usage of this CA will be compulsory for securing interacting with all banks of the country
      • Actually, this CA is not really a government entity, but a for-profit company that likes to make you pay through the nose
      • This government-sponsored monopoly likes to prop up other monopolies [microsoft.com] or create other monopolies [gemalto.com]
      • You'll be paying through the nose for gizmos such as signing sticks that don't act
  • by Zordak (123132) on Friday January 02, 2009 @10:22PM (#26307985) Homepage Journal
    Apparently somebody didn't get the memo that the only valid way to use this phrase anymore is to mock people who want to grow the enterprise by leveraging synergies.
  • by dencarl (138314) on Friday January 02, 2009 @10:41PM (#26308113)

    Why don't they use the method Google uses to verify control of a domain (and hence ownership)?

    The CA should require a unique file (containing a serial number) to be posted to a specific location on the website. Failing that you should be able to receive mail to an arbitrary email address at the domain.

    CAs who don't employ a technical measure (such as above) to verify domain ownership *prior* to issuing a cert would be taken out of the list of trusted CAs.

    • I believe StartCom and probably the other free providers do something like this. StartCom is in Firefox by default, by the way.

    • by cdrguru (88047)

      A brute force attack upon a server which gives you the ability to receive email through it or place files on it does not mean you have legal "control" over the domain.

      OK, it tends to indicate it but it is not any real assurance.

    • Re: (Score:3, Insightful)

      by blueg3 (192743)

      Kaminsky's DNS attack -- and the BGP hack, for that matter -- demonstrate pretty clearly why being able to masquerade as a particular host to the CA is not sufficient to prove you are actually the proper owner of that domain.

  • by lord_sarpedon (917201) on Friday January 02, 2009 @11:08PM (#26308317)

    Need a two tiered system.

    The world is so fucked up right now as far as censorship and snooping. We need encryption, everywhere, right now.

    Tier 1:
    "httpe" that acts similar to SSH - big warning on key changes. Known key can be included in html links even from untrusted sites (such as from a google search results page) for a cautionary warning with no loss of security. No prompt for a new site. Prompt if it changes. Prompt if a link gives a 'known' key different from the given one.

    Very easy to gradually deploy.

    Tier 2:
    Well-known certs for the root nameservers. Stick self-signed cert in DNS records. Sign DNS responses. Imposes a chain of trust type requirement on lesser nameservers.

    Tier 3:
    The fancier certs being passed around these days which are supposedly hyper deluxe verified. Actual monetary cost involved here. Determine a magic solution to make at least a few of the CAs trustworthy.

    • You are totally paranoid to a ridiculous degree (seriously, when was the last time you were censored, or even snooped on?), but you make a good point. It would be excellent if you could install your public key in the DNS server. Then if all traffic from the DNS server were encrypted, it would be extremely difficult to create a man-in-the-middle attack, in fact a number of attacks would be made quite difficult or impossible. There may even be a provision for this already in the DNS specification, since it
  • by Gothmolly (148874) on Friday January 02, 2009 @11:18PM (#26308363)

    The "industry" provided no value - it merely allowed you to pretend you were somehow secure, above and beyond the actual SSL part. Smoke and mirrors. If this "industry" dies, it will be a market correction, nothing more.

  • Bruce is wrong (Score:5, Insightful)

    by dachshund (300733) on Friday January 02, 2009 @11:29PM (#26308465)

    "SSL protects data in transit but the problem isn't eavesdropping on the transmission. Someone can steal the credit card on some server somewhere. The real risk is data in storage. SSL protects against the wrong problem," [Schneier] said.

    I respect Bruce, but I think if you say something true enough times, you lose sight of the fact that in this case it may not actually be a valid point. While credit card theft is a major problem, Phishers frequently target bank account login credentials--- which are not stored all over the place. In this case, SSL is one of the primary protections keeping you from all kind of hell (losing your credit card is a pain in the butt, but usually it's insured... losing your banking credentials can be a huge disaster). Now imagine that instead of a few rubes being conned by Phishing emails, you had millions of relatively savvy customers at a large ISP diverted to a fake Bank of America site (perhaps with help from insiders at the ISP). The losses could be substantial.

    Again, Bruce is right about one problem but not necessarily about every problem (and I can't help but notice that he works for a storage company...)

    • by jd (1658)

      Well, no, they're not stored every place. Usually, they're stored on the user's web browser or in some other similar system. As I recall from the paper on The Internet Auditing Project, their SSH security was broken because someone had the password on their Windows box and the Windows box was broken into. Also bear in mind that there were many stories in 2008 of servers being cracked, leading to the loss of hundreds of thousands, occasionally millions, of credit card numbers. So whilst I agree with you that

    • If counterpane is a storage company, then microsoft is a furniture company.
    • Re: (Score:3, Insightful)

      by blueg3 (192743)

      Actually, it's mostly popular to get bank credentials directly from the user's machine via malware. Jacking SSL isn't as successful.

      • > Jacking SSL isn't as successful.

        Yet.

      • "Jacking SSL isn't as successful."

        Jacking SSL wasn't sucessful exactly because it was strong, so there was no known vector to attack it. Since malware couldn't attack SSL, they refrained to less efficient tatics, like relying on the ignorance of the user.

        Now, that SSL is broken, it is almost certain that they will start to use the more efficient attacks, that directly target SSL.

        • by blueg3 (192743)

          SSL isn't really broken. There's one attack against it. It was already known that the attack was possible, the only question was the difficulty. As it stands, reproducing their work is fairly difficult. There's also a quite effective mitigation -- don't accept certificates where any elements of the certificate chain other than pre-trusted root certs use only MD5 as their hash algorithm.

  • by Animats (122034) on Saturday January 03, 2009 @12:02AM (#26308677) Homepage

    There are really three tiers of SSL certs being sold:

    1. "Domain control only validated" certs. This means the cert issuer got an answer from an e-mail sent to the domain. This is the "QuickSSL" tier.
    2. "Location and business identiti validated" certs. What SSL certs were supposed to mean. The cert issuer actually checked out the business for existence. At this tier, there's often a "relying party" guarantee.
    3. "Extended validation" certs. The cert issuer had to meet some audited standards to issue the cert. Mostly used by banks.

    Current browsers don't distinguish between #1 and #2. They should. "Domain control only validated" certs are enough to secure some social networking site or blog, but not good enough to send someone a credit card number. If they're taking your money, the cert should contain enough info to allow you to find and sue them.

    Our SiteTruth [sitetruth.com] system distinguishes between #1 and #2, because we're looking for business identity. It's a useful way to filter out the "bottom feeders".

    The problems with bogus SSL cert issuance seem to be, so far, confined to the "Domain control only validated" certs. This is an additional good reason to distinguish between them and the better tiers.

    • Re: (Score:3, Interesting)

      by sjames (1099)

      Personally, I lost faith in the CAs and the certs they sign early on. I was at a sort of b2b expo (The dot-com boom was just barely beginning but nobody knew it).

      I met a representative from a CA that I won't identify, but I'm sure you've heard of them. He came prepared to give 'why you need a cert and https' sales pitch to various sorts of people from CEO to sales to CTO to techie.

      He wasn't (apparently) prepared to discuss trust and authentication in any depth. When he told me (paraphrased) that they "KNOW

      • When he told me (paraphrased) that they "KNOW the entity they give a cert to isn't committing fraud because they have to sign a LEGAL DOCUMENT that says they aren't!"

        A marketroid spouting nonsense about technical matters. What else is new?

        Of course, you and I know that a CA is supposed to verify identity of the party that they're issuing a certificate to, not its trustworthiness (unless they're issuing a sub-CA certificate, but that's a different matter). Much misunderstanding does indeed come from this misconception of a CA's role.

        Of course, https is screwed up anyway because of the way it munges security and authenticity together. Ideally, browser and server should immediately do a key exchange, then once the connection is encrypted, perform optional authentication after the browser sends the host field. The lock icon should indicate encryption and authentication separately.

        Ok, now you seem to fall prey to the same misconception. Without being sure about the identity of the party your communicating with, there

        • by sjames (1099)

          Actually, I am well aware of the issues w/ MITM attacks. However, IF ssl established encryption, then virtual host, then authenticity (to varying degrees of confidence as configured by the user's trust settings), https wouldn't require a seperate IP per virtual host anymore. If I'm talking w/ my bank, I'll surely require authentication as well as encryption.

          There are cases (such as a private lan to an internal server) where authentication (particularly CLIENT authentication) w/o encryption may be reasonable

          • However, IF ssl established encryption, then virtual host, then authenticity (to varying degrees of confidence as configured by the user's trust settings), https wouldn't require a seperate IP per virtual host anymore.

            It doesn't. There are several approaches to the problem:

            • subjectAltNames [robichaux.net] and wildcard certificates. Nowadays, certificates can carry multiple site names. Just make a certificate containing all site names (not supported by all CA's, but Entrust and CaCert [cacert.org] do)
            • SNI (Server Name Indication) [wikipedia.org]. This allows the webserver to use a different certificate for different virtual hosts. Soon to be supported by mod_ssl, but already supported (for several years) by mod_gnutls [outoforder.cc]
            • . Most clients (browsers) today support it out

Whenever a system becomes completely defined, some damn fool discovers something which either abolishes the system or expands it beyond recognition.

Working...