Re: security questions
[EMAIL PROTECTED] wrote: John Ioannidis wrote: | Does anyone know how this security questions disease started, and why | it is spreading the way it is? If your company does this, can you find | the people responsible and ask them what they were thinking? The answer is Help Desk Call Avoidance; allow the end-user to fix their own account without having to get someone on the phone. This is simply an available mechanism in the spectrum between easy-to-use and rock-solid security. As the discussion so far indicates, and as published papers show, the security of these security questions is lower than the security of the password. | My theory is that no actual security people have ever been involved, and | that it's just another one of those stupid design practices that are | perpetuated because nobody has ever complained or that's what | everybody is doing. Your theory is incorrect. There is considerable analysis on what Can you reference it please? There has been some analysis on the entropy of passphrases as a password replacement, but it is not relevant. constitute good security questions based on the anticipated entropy of the responses. This is why, for example, no good security question has a yes/no answer (i.e., 1-bit). Aren't security questions just an automation of what happens once you get a customer service representative on the phone? In some regards they may be more secure as they're less subject to social manipulation (i.e., if I mention a few possible answers to a customer support person, I can probably get them to confirm an answer for me). The difference is that when you are interfacing with a human, you have to go through a low-speed interface, namely, voice. In that respect, a security question, coupled with a challenge about recent transactions, makes for adequate security. The on-line version of the security question is vulnerable to automated dictionary attacks. /ji - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
More man-in-the-middle'd SSL sessions on the way
From an article about WAN optimization appliances in Computerworld: In some markets, such as health and finance, [hiring] a managed provider [who will do the encryption outside your routers] isn't a good option for another reason: Because data is optimized in an unencrypted state, privacy and security concerns arise. But vendors such as Riverbed, Juniper Networks and Blue Coat Systems can serve as a trusted man in the middle for optimizing data encrypted with SSL, which is commonly used in applications with Web interfaces and other Internet traffic. They terminate the encrypted session, decrypt, optimize and then re-encrypt and forward the traffic. [Gartner's Joe] Skorupa said most vendors are developing this useful capability. It may indeed be a useful capability - but widespread use will destroy what little is left of the SSL trust model. -- Jerry - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
OpenID/Debian PRNG/DNS Cache poisoning advisory
Security Advisory (08-AUG-2008) (CVE-2008-3280) === Ben Laurie of Google's Applied Security team, while working with an external researcher, Dr. Richard Clayton of the Computer Laboratory, Cambridge University, found that various OpenID Providers (OPs) had TLS Server Certificates that used weak keys, as a result of the Debian Predictable Random Number Generator (CVE-2008-0166). In combination with the DNS Cache Poisoning issue (CVE-2008-1447) and the fact that almost all SSL/TLS implementations do not consult CRLs (currently an untracked issue), this means that it is impossible to rely on these OPs. Attack Description -- In order to mount an attack against a vulnerable OP, the attacker first finds the private key corresponding to the weak TLS certificate. He then sets up a website masquerading as the original OP, both for the OpenID protocol and also for HTTP/HTTPS. Then he poisons the DNS cache of the victim to make it appear that his server is the true OpenID Provider. There are two cases, one is where the victim is a user trying to identify themselves, in which case, even if they use HTTPS to ensure that the site they are visiting is indeed their provider, they will be unable to detect the substitution and will give their login credentials to the attacker. The second case is where the victim is the Relying Party (RP). In this case, even if the RP uses TLS to connect to the OP, as is recommended for higher assurance, he will not be defended, as the vast majority of OpenID implementations do not check CRLs, and will, therefore, accept the malicious site as the true OP. Mitigation -- Mitigation is surprisingly hard. In theory the vulnerable site should revoke their weak certificate and issue a new one. However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. 2. OpenID libraries must check CRLs. 3. DNS caching resolvers must be patched against the poisoning attack. 4. Until either 1 and 2 or 3 have been done, OpenID cannot be trusted for any OP that cannot demonstrate it has never had a weak certificate. Discussion -- Normally, when security problems are encountered with a single piece of software, the responsible thing to do is to is to wait until fixes are available before making any announcement. However, as a number of examples in the past have demonstrated, this approach does not work particularly well when many different pieces of software are involved because it is necessary to coordinate a simultaneous release of the fixes, whilst hoping that the very large number of people involved will cooperate in keeping the vulnerability secret. In the present situation, the fixes will involve considerable development work in adding CRL handling to a great many pieces of openID code. This is a far from trivial amount of work. The fixes will also involve changes to browser preferences to ensure that CRLs are checked by default -- which many vendors have resisted for years. We are extremely pessimistic that a security vulnerability in OpenID will be seen as sufficiently important to change the browser vendors minds. Hence, we see no value in delaying this announcement; and by making the details public as soon as possible, we believe that individuals who rely on OpenID will be better able to take their own individual steps to avoid relying upon the flawed certificates we have identified. OpenID is at heart quite a weak protocol, when used in its most general form[1], and consequently there is very limited reliance upon its security. This means that the consequences of the combination of attacks that are now possible is nothing like as serious as might otherwise have been the case. However, it does give an insight into the type of security disaster that may occur in the future if we do not start to take CRLs seriously, but merely stick them onto to-do lists or disable them in the name of tiny performance improvements. Affected Sites -- There is no central registry of OpenID systems, and so we cannot be sure that we have identified all of the weak certificates that are currently being served. The list of those we have found so far is: openid.sun.com www.xopenid.net openid.net.nz Notes - [1] There are ways of using OpenID that are significantly more secure than the commonly deployed scheme, I shall describe those in a separate article. - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: security questions
| | My theory is that no actual security people have ever been involved, | | that it's just another one of those stupid design practices that are | | perpetuated because nobody has ever complained or that's what | | everybody is doing. | | Your theory is incorrect. There is considerable analysis on what | | Can you reference it please? There has been some analysis on the | entropy of passphrases as a password replacement, but it is not | relevant. RSA sells a product that is based on such research. I don't have references; perhaps someone else does. I think the accurate statement here is: There's been some research on this matter, and there are some reasonable implementations out there; but there are also plenty of me-too implementations that are quite worthless. In fact, I've personally never run into an implementation that I would not consider worthless. (Oddly, the list of questions that started this discussion is one of the better ones I've seen. Unfortunately, what it demonstrates is that producing a useful implementation with a decent amount of total entropy probably involves more setup time than the average user will want to put up with.) | constitute good security questions based on the anticipated entropy | of the responses. This is why, for example, no good security | question has a yes/no answer (i.e., 1-bit). Aren't security | questions just an automation of what happens once you get a customer | service representative on the phone? In some regards they may be | more secure as they're less subject to social manipulation (i.e., if | I mention a few possible answers to a customer support person, I can | probably get them to confirm an answer for me). | The difference is that when you are interfacing with a human, you have | to go through a low-speed interface, namely, voice. In that respect, a | security question, coupled with a challenge about recent transactions, | makes for adequate security. The on-line version of the security | question is vulnerable to automated dictionary attacks. Actually, this cuts both ways. Automated interfaces generally require exact matches; at most, they will be case-blind. This is appropriate and understood for passwords. It is inappropriate for what people perceive as natural-text questions and answers. When I first started running into such systems, when asked for where I was born, I would answer New York - or maybe New York City, or maybe NY or NYC. I should have thought about the consequences of providing a natural- text answer to a natural-text question - but I didn't. Sure enough, when I actually needed to reset my password - I ended up getting locked out of the system because there was no way I could remember, 6 months later, what exact answer I'd given. A human being is more forgiving. This makes the system more vulnerable to social engineering - but it makes it actually useable. The tradeoff here is very difficult to make. By its nature, a secondary access system will be rarely used. People may, by dint of repetition, learn to parrot back exact answers, even a random bunch of characters, if they have to use them every day. There's no way anything but a fuzzy match on meaning will work for an answer people have to give once every couple of months - human memory simply doesn't work that way. I learned my lesson and never provide actual answers to these questions any more. -- Jerry - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
At Fri, 8 Aug 2008 11:50:59 +0100, Ben Laurie wrote: However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. Isn't this a good argument for blacklisting the keys on the client side? -Ekr - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
UK e-passport cloned
Quoting: New microchipped passports designed to be foolproof against identity theft can be cloned and manipulated in minutes and accepted as genuine by the computer software recommended for use at international airports. Tests for The Times exposed security flaws in the microchips introduced to protect against terrorism and organised crime. The flaws also undermine claims that 3,000 blank passports stolen last week were worthless because they could not be forged. http://www.timesonline.co.uk/tol/news/uk/crime/article4467106.ece Hat tip: Bruce Schneier's blog -- Perry E. Metzger[EMAIL PROTECTED] - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
RE: OpenID/Debian PRNG/DNS Cache poisoning advisory
Eric Rescorla wrote on 08 August 2008 16:06: At Fri, 8 Aug 2008 11:50:59 +0100, Ben Laurie wrote: However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. Isn't this a good argument for blacklisting the keys on the client side? Isn't that exactly what Browsers must check CRLs means in this context anyway? What alternative client-side blacklisting mechanism do you suggest? cheers, DaveK -- Can't think of a witty .sigline today - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
At Fri, 8 Aug 2008 17:31:15 +0100, Dave Korn wrote: Eric Rescorla wrote on 08 August 2008 16:06: At Fri, 8 Aug 2008 11:50:59 +0100, Ben Laurie wrote: However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. Isn't this a good argument for blacklisting the keys on the client side? Isn't that exactly what Browsers must check CRLs means in this context anyway? What alternative client-side blacklisting mechanism do you suggest? It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... -Ekr - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
RE: OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, 8 Aug 2008, Dave Korn wrote: | Isn't this a good argument for blacklisting the keys on the client | side? | | Isn't that exactly what Browsers must check CRLs means in this | context anyway? What alternative client-side blacklisting mechanism | do you suggest? Since the list of bad keys is known and fairly short, one could explicitly check for them in the browser code, without reference to any external CRL. Of course, the browser itself may not see the bad key - it may see key for something that *contains* a bad key. So such a check would not be complete. Still, it couldn't hurt. One could put similar checks everywhere that keys are used. Think of it as the modern version of code that checks for and rejects DES weak and semi-weak keys. The more code out there that does the check, the faster bad keys will be driven out of use. -- Jerry - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
RE: OpenID/Debian PRNG/DNS Cache poisoning advisory
Eric Rescorla wrote on 08 August 2008 17:58: At Fri, 8 Aug 2008 17:31:15 +0100, Dave Korn wrote: Eric Rescorla wrote on 08 August 2008 16:06: At Fri, 8 Aug 2008 11:50:59 +0100, Ben Laurie wrote: However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. Isn't this a good argument for blacklisting the keys on the client side? Isn't that exactly what Browsers must check CRLs means in this context anyway? What alternative client-side blacklisting mechanism do you suggest? It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... scurries off to read CRL format in RFC Oh, you can't specify them solely by key, you have to have all the associated metadata. That's annoying, yes, I understand your point now. IIRC various of the vendors' sshd updates released in the immediate wake of the Debian catastrophe do indeed block all the weak keys. cheers, DaveK -- Can't think of a witty .sigline today - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, Aug 8, 2008 at 5:57 PM, Eric Rescorla [EMAIL PROTECTED] wrote: At Fri, 8 Aug 2008 17:31:15 +0100, Dave Korn wrote: Eric Rescorla wrote on 08 August 2008 16:06: At Fri, 8 Aug 2008 11:50:59 +0100, Ben Laurie wrote: However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. Isn't this a good argument for blacklisting the keys on the client side? Isn't that exactly what Browsers must check CRLs means in this context anyway? What alternative client-side blacklisting mechanism do you suggest? It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... It also only fixes this single type of key compromise. Surely it is time to stop ignoring CRLs before something more serious goes wrong? - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
Eric Rescorla wrote: At Fri, 8 Aug 2008 17:31:15 +0100, Dave Korn wrote: Eric Rescorla wrote on 08 August 2008 16:06: At Fri, 8 Aug 2008 11:50:59 +0100, Ben Laurie wrote: However, since the CRLs will almost certainly not be checked, this means the site will still be vulnerable to attack for the lifetime of the certificate (and perhaps beyond, depending on user behaviour). Note that shutting down the site DOES NOT prevent the attack. Therefore mitigation falls to other parties. 1. Browsers must check CRLs by default. Isn't this a good argument for blacklisting the keys on the client side? Isn't that exactly what Browsers must check CRLs means in this context anyway? What alternative client-side blacklisting mechanism do you suggest? It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... Funnily enough I was just working on this -- and found that we'd end up adding a couple megabytes to every browser. #DEFINE NONSTARTER. I am curious about the feasibility of a large bloom filter that fails back to online checking though. This has side effects but perhaps they can be made statistically very unlikely, without blowing out the size of a browser. Updating the filter could then be something we do on a 24 hour autoupdate basis. Doing either this, or doing revocation checking over DNS (seriously), is not necessarily a bad idea. We need to do better than we've been. - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
Eric Rescorla [EMAIL PROTECTED] writes: It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... You'd also end up with a rather large list for the client to carry around, which would be especially problematic for lightweight clients. You'd need to represent it as something like a Bloom filter to avoid this (given that most users will just click OK on invalid certs, the small false positive rate shouldn't have much effect either :-). Peter. - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
Ben Laurie [EMAIL PROTECTED] writes: It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... It also only fixes this single type of key compromise. Surely it is time to stop ignoring CRLs before something more serious goes wrong? The problem is, the CRL mechanism itself is also dangerous. Sadly, clients are required to keep on going if they can't reach a CRL server. That means that if you DoSing the CRL servers or use DNS attacks to effectively take them offline, you've also effectively eliminated the certificate revocation. I'm not going to tell you that paying attention to CRLs wouldn't be better than what happens now, but it will not eliminate the problem. It is too hard to prove a negative (that is, to prove to yourself that no revocation exists.) The kerberos style of having credentials expire very quickly is one (somewhat less imperfect) way to deal with such things, but it is far from perfect and it could not be done for the ad-hoc certificate system https: depends on -- the infrastructure for refreshing all the world's certs every eight hours doesn't exist, and if it did imagine the chaos if it failed for a major CA one fine morning. One also worries about what will happen in the UI when a certificate has been revoked. If it just says this cert has been revoked, continue anyway? the wrong thing will almost always happen. Perry -- Perry E. Metzger[EMAIL PROTECTED] - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
At Fri, 08 Aug 2008 10:43:53 -0700, Dan Kaminsky wrote: Eric Rescorla wrote: It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... Funnily enough I was just working on this -- and found that we'd end up adding a couple megabytes to every browser. #DEFINE NONSTARTER. I am curious about the feasibility of a large bloom filter that fails back to online checking though. This has side effects but perhaps they can be made statistically very unlikely, without blowing out the size of a browser. Why do you say a couple of megabytes? 99% of the value would be 1024-bit RSA keys. There are ~32,000 such keys. If you devote an 80-bit hash to each one (which is easily large enough to give you a vanishingly small false positive probability; you could probably get away with 64 bits), that's 320KB. Given that the smallest Firefox build (Windows) is 7.1 MB, this doesn't sound like a nonstarter to me at all, especially since the browser could download it in the background. Updating the filter could then be something we do on a 24 hour autoupdate basis. Doing either this, or doing revocation checking over DNS (seriously), is not necessarily a bad idea. We need to do better than we've been. Yes, there are a number of approaches to more efficient CRL checking, I think that's a separate issue. -Ekr - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, Aug 08, 2008 at 02:08:37PM -0400, Perry E. Metzger wrote: The kerberos style of having credentials expire very quickly is one (somewhat less imperfect) way to deal with such things, but it is far from perfect and it could not be done for the ad-hoc certificate system https: depends on -- the infrastructure for refreshing all the world's certs every eight hours doesn't exist, and if it did imagine the chaos if it failed for a major CA one fine morning. The PKIX moral equivalent of Kerberos V tickets would be OCSP Responses. I understand most current browsers support OCSP. One also worries about what will happen in the UI when a certificate has been revoked. If it just says this cert has been revoked, continue anyway? the wrong thing will almost always happen. No doubt. - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, Aug 08, 2008 at 11:20:15AM -0700, Eric Rescorla wrote: At Fri, 08 Aug 2008 10:43:53 -0700, Dan Kaminsky wrote: Funnily enough I was just working on this -- and found that we'd end up adding a couple megabytes to every browser. #DEFINE NONSTARTER. I am curious about the feasibility of a large bloom filter that fails back to online checking though. This has side effects but perhaps they can be made statistically very unlikely, without blowing out the size of a browser. Why do you say a couple of megabytes? 99% of the value would be 1024-bit RSA keys. There are ~32,000 such keys. If you devote an 80-bit hash to each one (which is easily large enough to give you a vanishingly small false positive probability; you could probably get away with 64 bits), that's 320KB. Given that the smallest Firefox [...] You could store {hash, seed} and check matches for false positives by generating a key with the corresponding seed and then checking for an exact match -- slow, but rare. This way you could choose your false positive rate / table size comfort zone and vary the size of the hash accordingly. Nico -- - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
At 1:47 PM -0500 8/8/08, Nicolas Williams wrote: On Fri, Aug 08, 2008 at 02:08:37PM -0400, Perry E. Metzger wrote: The kerberos style of having credentials expire very quickly is one (somewhat less imperfect) way to deal with such things, but it is far from perfect and it could not be done for the ad-hoc certificate system https: depends on -- the infrastructure for refreshing all the world's certs every eight hours doesn't exist, and if it did imagine the chaos if it failed for a major CA one fine morning. The PKIX moral equivalent of Kerberos V tickets would be OCSP Responses. I understand most current browsers support OCSP. ...and only a tiny number of CAs do so. --Paul Hoffman, Director --VPN Consortium - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Telephone Phishing
I just got called by an autodialer -- the Caller ID was faked (and in any case didn't point at a real number since area codes don't start with 0 -- probably a mistake by the scammers). After I answered, a tape of a cheerful woman informed me this was my last chance to lower the rate on my credit card, and asked me to press one to continue. You can fill in the rest of the script on your own. I'm sure this happens all the time now and I was just unaware of it, but it is always more vivid when you see it yourself. I'm certain this scenario would get enough average people to hand over their credit card data to more than pay for itself, and smart scammers are probably using VOIP accounts they got with stolen credit card numbers to do this anyway. One can also imagine using this technique for a wide variety of spear phishing attacks. For example, say you stole a large number of credit card numbers but didn't have the CVV2s -- you could set up an IVR system to automatically collect them from your victims. -- Perry E. Metzger[EMAIL PROTECTED] - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: [OpenID] OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, Aug 8, 2008 at 8:27 PM, Eddy Nigg (StartCom Ltd.) [EMAIL PROTECTED] wrote: Ben Laurie: On Fri, Aug 8, 2008 at 12:44 PM, Eddy Nigg (StartCom Ltd.) [EMAIL PROTECTED] wrote: This affects any web site and service provider of various natures. It's not exclusive for OpenID nor for any other protocol / standard / service! It may affect an OpenID provider if it uses a compromised key in combination with unpatched DNS servers. I don't understand why OpenID is singled out, since it can potentially affect any web site including Google's various services (if Google would have used Debian systems to create their private keys). OpenID is singled out because I am not talking about a potential problem but an actual problem. Sorry Ben, but any web site or service (HTTP, SMPT, IMAP, SSH, VPN, etc) which makes use of a compromised key has an actual problem and not a potential problem. Open ID as a standard isn't more affected than, lets say XMPP...If there are servers and providers relying on such keys the have a real actual problem. I do not dispute this. I don't see your point about Open ID nor didn't I see anything new The point is I found OpenID servers with weak keys. The problem of weak keys should be dealt at the CA level, many which have failed to do anything serious about it. Indeed. We have spotted other actual problems in other services. Details will be forthcoming at appropriate times. I think it's superfluous to single out different services since any service making use of the weak keys is affected, with recent discovery of DNS poisoning making the matter worse. I suggest you try a forum which can potentially reach many CAs, they in fact have everything at their disposal to remove this threat! If you have a better forum, bring it on. However, CAs do not have everything at their disposal to remove the threat. Browsers,OpenID libraries and RPs must also participate. Just as saying buffer overflows are bad has not magically caused all buffer overflows to be fixed, I am confident that the only way to get this problem fixed is to chase down all the culprits individually. I am sure that OpenID is not the only thing with problems, as you say. - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
| Funnily enough I was just working on this -- and found that we'd | end up adding a couple megabytes to every browser. #DEFINE | NONSTARTER. I am curious about the feasibility of a large bloom | filter that fails back to online checking though. This has side | effects but perhaps they can be made statistically very unlikely, | without blowing out the size of a browser. | Why do you say a couple of megabytes? 99% of the value would be | 1024-bit RSA keys. There are ~32,000 such keys. If you devote an | 80-bit hash to each one (which is easily large enough to give you a | vanishingly small false positive probability; you could probably get | away with 64 bits), that's 320KB. Given that the smallest Firefox | [...] You can get by with a lot less than 64 bits. People see problems like this and immediately think birthday paradox, but there is no birthday paradox here: You aren't look for pairs in an ever-growing set, you're looking for matches against a fixed set. If you use 30-bit hashes - giving you about a 120KB table - the chance that any given key happens to hash to something in the table is one in a billion, now and forever. (Of course, if you use a given key repeatedly, and it happens to be that 1 in a billion, it will hit every time. So an additional table of known good keys that happen to collide is worth maintaining. Even if you somehow built and maintained that table for all the keys across all the systems in the world - how big would it get, if only 1 in a billion keys world-wide got entered?) | You could store {hash, seed} and check matches for false positives | by generating a key with the corresponding seed and then checking for an | exact match -- slow, but rare. This way you could choose your false | positive rate / table size comfort zone and vary the size of the hash | accordingly. Or just go off to one of a number of web sites that have a full table. Many solutions are possible, when they only need to be invoked very, very rarely. -- Jerry | Nico | -- | | - | The Cryptography Mailing List | Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED] | | - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
[Sorry for duplicates, but I got multiple requests for a non-HTML version, and I didn't want to fork the thread. Also sorry for initially sending HTML; I didn't realize it was so abhorrent these days. ] On Fri, Aug 8, 2008 at 1:43 PM, Dan Kaminsky [EMAIL PROTECTED] wrote: It's easy to compute all the public keys that will be generated by the broken PRNG. The clients could embed that list and refuse to accept any certificate containing one of them. So, this is distinct from CRLs in that it doesn't require knowing which servers have which cert... Funnily enough I was just working on this -- and found that we'd end up adding a couple megabytes to every browser. #DEFINE NONSTARTER. I am curious about the feasibility of a large bloom filter that fails back to online checking though. This has side effects but perhaps they can be made statistically very unlikely, without blowing out the size of a browser. Using this Bloom filter calculator: http://www.cc.gatech.edu/~manolios/bloom-filters/calculator.html , plus the fact that there are 32,768 weak keys for every key type size, I get various sizes of necessary Bloom filter, based on how many key type / sizes you want to check and various false positive rates: * 3 key types/sizes with 1e-6 false positive rate: 2826759 bits = 353 KB * 3 key types/sizes with 1e-9 false positive rate: 4240139 bits = 530 KB * 7 key types/sizes with 1e-6 false positive rate: 6595771 bits = 824 KB * 7 key types/sizes with 1e-9 false positive rate: 9893657 bits = 1237 KB I presume that the first 3 first 7 key type/sizes in this list http://metasploit.com/users/hdm/tools/debian-openssl/ are the best to incorporate into the filter. Is there any chance it would be feasible to get a list of all the weak keys that were actually certified by browser-installed CAs, or those weak certificates? Presumably, this list would be much smaller and would be more effectively distributed in Bloom filter form. - Tim - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, Aug 8, 2008 at 7:54 PM, Tim Dierks [EMAIL PROTECTED] wrote: Using this Bloom filter calculator: http://www.cc.gatech.edu/~manolios/bloom-filters/calculator.html , plus the fact that there are 32,768 weak keys for every key type size, I get various sizes of necessary Bloom filter, based on how many key type / sizes you want to check and various false positive rates: * 3 key types/sizes with 1e-6 false positive rate: 2826759 bits = 353 KB * 3 key types/sizes with 1e-9 false positive rate: 4240139 bits = 530 KB * 7 key types/sizes with 1e-6 false positive rate: 6595771 bits = 824 KB * 7 key types/sizes with 1e-9 false positive rate: 9893657 bits = 1237 KB I presume that the first 3 first 7 key type/sizes in this list http://metasploit.com/users/hdm/tools/debian-openssl/ are the best to incorporate into the filter. Is there any chance it would be feasible to get a list of all the weak keys that were actually certified by browser-installed CAs, or those weak certificates? Presumably, this list would be much smaller and would be more effectively distributed in Bloom filter form. Or as a CRL :-) - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
On Fri, Aug 08, 2008 at 12:35:43PM -0700, Paul Hoffman wrote: At 1:47 PM -0500 8/8/08, Nicolas Williams wrote: On Fri, Aug 08, 2008 at 02:08:37PM -0400, Perry E. Metzger wrote: The kerberos style of having credentials expire very quickly is one (somewhat less imperfect) way to deal with such things, but it is far from perfect and it could not be done for the ad-hoc certificate system https: depends on -- the infrastructure for refreshing all the world's certs every eight hours doesn't exist, and if it did imagine the chaos if it failed for a major CA one fine morning. The PKIX moral equivalent of Kerberos V tickets would be OCSP Responses. I understand most current browsers support OCSP. ...and only a tiny number of CAs do so. Not that long ago nothing supported OCSP. If all that's left (ha) is the CAs then we're in good shape. (OCSP services can be added without modifying a CA -- just issue the OCSP Responders their certs and let them use CRLs are their source of revocation information.) - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
At Fri, 8 Aug 2008 15:52:07 -0400 (EDT), Leichter, Jerry wrote: | Funnily enough I was just working on this -- and found that we'd | end up adding a couple megabytes to every browser. #DEFINE | NONSTARTER. I am curious about the feasibility of a large bloom | filter that fails back to online checking though. This has side | effects but perhaps they can be made statistically very unlikely, | without blowing out the size of a browser. | Why do you say a couple of megabytes? 99% of the value would be | 1024-bit RSA keys. There are ~32,000 such keys. If you devote an | 80-bit hash to each one (which is easily large enough to give you a | vanishingly small false positive probability; you could probably get | away with 64 bits), that's 320KB. Given that the smallest Firefox | [...] You can get by with a lot less than 64 bits. People see problems like this and immediately think birthday paradox, but there is no birthday paradox here: You aren't look for pairs in an ever-growing set, you're looking for matches against a fixed set. If you use 30-bit hashes - giving you about a 120KB table - the chance that any given key happens to hash to something in the table is one in a billion, now and forever. (Of course, if you use a given key repeatedly, and it happens to be that 1 in a billion, it will hit every time. So an additional table of known good keys that happen to collide is worth maintaining. Even if you somehow built and maintained that table for all the keys across all the systems in the world - how big would it get, if only 1 in a billion keys world-wide got entered?) I don't believe your math is correct here. Or rather, it would be correct if there was only one bad key. Remember, there are N bad keys and you're using a b-bit hash, which has 2^b distinct values. If you put N' entries in the hash table, the probability that a new key will have the same digest as one of them is N'/(2^b). If b is sufficiently large to make collisions rare, then N'=~N and we get N/(2^b). To be concrete, we have 2^15 distinct keys, so, the probability of a false positive becomes (2^15)/(2^b)=2^(b-15). To get that probability below 1 billion, b+15 = 30, so you need about 45 bits. I chose 64 because it seemed to me that a false positive probability of 2^{-48} or so was better. -Ekr - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
| You can get by with a lot less than 64 bits. People see problems | like this and immediately think birthday paradox, but there is no | birthday paradox here: You aren't look for pairs in an | ever-growing set, you're looking for matches against a fixed set. | If you use 30-bit hashes - giving you about a 120KB table - the | chance that any given key happens to hash to something in the table | is one in a billion, now and forever. (Of course, if you use a | given key repeatedly, and it happens to be that 1 in a billion, it | will hit every time. So an additional table of known good keys | that happen to collide is worth maintaining. Even if you somehow | built and maintained that table for all the keys across all the | systems in the world - how big would it get, if only 1 in a billion | keys world-wide got entered?) | I don't believe your math is correct here. Or rather, it would | be correct if there was only one bad key. | | Remember, there are N bad keys and you're using a b-bit hash, which | has 2^b distinct values. If you put N' entries in the hash table, the | probability that a new key will have the same digest as one of them is | N'/(2^b). If b is sufficiently large to make collisions rare, then | N'=~N and we get N/(2^b). | | To be concrete, we have 2^15 distinct keys, so, the probability of a | false positive becomes (2^15)/(2^b)=2^(b-15). To get that probability | below 1 billion, b+15 = 30, so you need about 45 bits. I chose 64 | because it seemed to me that a false positive probability of 2^{-48} | or so was better. You're right, of course - I considered 32,000 to be vanishingly small compared to the number of hash values, but of course it isn't. The perils of looking at one number just as decimal and the other just in exponential form In any case, I think it's clear that even for extremely conservative false hit ratios, the table size is quite reasonable. You wouldn't want the table on your smart card or RFID chip, perhaps, but there even a low-end smartphone would have no problems. -- Jerry - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
* Eric Rescorla: Why do you say a couple of megabytes? 99% of the value would be 1024-bit RSA keys. There are ~32,000 such keys. There are three sets of keys, for big-endian 32-bit, little-endian 32-bit and little-endian 64-bit. On top of that, openssl genrsa generates different keys depending on the existence of $HOME/.rnd (and -3 creates yet another set of keys, but this is more in the league of different key length). If the library is used for key generation (instead of the command line tool), different keys might result. On the other hand, the on-disk size would be comparable to the phishing filter database. Part of the problem of the CRL approach is that CAs usually have policies against obtaining private keys and therefore can't prove to the customer that their keys are compromised. And adding a CRL entry when the customer isn't convinced that they've got a problem is probably not a good idea, either. - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]
Re: OpenID/Debian PRNG/DNS Cache poisoning advisory
[I feel a little uncomfortable replying with such a wide distribution!] Getting browsers, or OpenID installations, to check CRLs or use OCSP to check for freshness is likely to be slow going. At this point I think the momentum still favors fixing the remaining DNS systems that are vulnerable to cache poisoning. This turnkey-MITM bug makes OpenSSL bad certs far more exploitable, as Dan Kaminsky pointed out in his report. OpenID is just one example of many where this is going to keep happening as long as DNS is unpatched. I thought of one possible mitigation that can protect OpenID end users against remote web sites which have not patched their DNS. OpenID providers who used weak OpenSSL certs would have to change their URLs so that their old X.509 CA certs on their old URLs no longer work on the new ones. This will require all of their clients (users who log in with their OpenID credentials) to change their identifiers. DNS based MITMs will not be able to forge messages related to the new identifiers. Customers can be alerted to this requirement as soon as they log in to a web site (relying party) whose DNS is NOT hacked; the redirection to the OpenID provider will give opportunity to notify the customer of the name change. Making this change may be somewhat inconvenient, but since OpenID is a relatively new standard, at least it is easier than would be the case with a more established protocol. In the other direction of attack, the end user's DNS is poisoned and he gets redirected to a bogus site in place of the OpenID provider; that site is then able to provide a valid SSL certificate due to the OpenSSL weakness, thereby stealing whatever authentication credentials the user normally sends to his OpenID provider. This is one instance of the general attack where a user is DNS-misdirected to a bogus copy of a secure site which unfortunately used weak OpenSSL based certs. Again, I see fixing the DNS as the path of least resistance here, especially so since the end user is the one bearing most of the risk, typically DNS is provided by an ISP or some other agency with a formal legal relationship, and there is the possibility of liability on the part of the lax DNS provider. Hopefully we will continue to see rapid uptake of the DNS fix over the next few weeks. That still leaves weak-cert OpenID users vulnerable to DNS-unpatched service providers (OpenID relying parties), and that is where my proposed mitigation above comes in. By renaming its URLs, an OpenID provider who had the misfortune to create a weak OpenSSL cert (through no fault of its own) can save its end users considerable potential grief. Hal Finney - The Cryptography Mailing List Unsubscribe by sending unsubscribe cryptography to [EMAIL PROTECTED]