Duncans Security Blog An enthusiasts musings


Heartbleed: my comments

It has been 11 days since the public disclosure of a major bug in OpenSSL, known as Heartbleed. I have been asked about my thoughts by a few people(both technical and non-technical), and so I find myself writing this blog post. I must mention that security disclosures occur on a weekly(if not, daily) basis, but heartbleed has been one that has stood out, even to laymen. To clear the air, there is a lot of discussion about stolen passwords. This isnt necessarily true. The passwords could potentially have been stolen, but it doesnt mean they have been(and unfortunately, there is no way to know if they have been compromised).

So what is it about? Is it important? What is being overlooked?

What is it, in a nutshell?
Heartbleed is the name that has been given to a bug which was found in OpenSSL. OpenSSL is an open source implementation of the SSL and TLS protocols, which are used for secure communications across a network. OpenSSL is also by far, the most widely used implementation of SSL/TLS in the world. If you have ever logged in to any website on the internet, there is a very good chance that OpenSSL was involved. Some of the major websites that use OpenSSL, are listed in the Heartbleed hit-list.

The bug revolves around a piece of code committed to OpenSSLs codebase in 2012, known as heartbeat. OpenSSL makes use of a process of secret exchanges in order to authenticate the client and server, in order to establish a secure connection. The heartbeat feature is used to keep existing sessions "alive", when there is no data being transmitted. Heartbeat involves an exchange of an arbitrary message from the client to the server, and the subsequent reply from the server, with the same arbitrary message. However, security researchers found that there was no bound check done on this arbitrary message. So they discovered that when receiving their arbitrary message back, they could request a bigger message in response, up to 64K in size.

This meant that the server would "bleed"(hence the name, "heartbleed") the extra bits of information that where not in the original message from heartbeat, and this additional data would come directly from the servers memory. This has been nicely summarised in a comic by XKCD. Due to the nature of heartbleed, there is unfortunately no way that the use of heartbleed could have been detected, and so it is not known if or where the exploit has been used.

What can be bled from the servers memory?
The spoils are random dumps of the next blocks of memory after the heartbeat message. So this may or may not be of use to an attacker. However, it is entirely possible that the data could consist (either in full, or in part) of sensitive information, such as; the servers private key(the holy grail), user names and passwords, credit card information, sensitive documents, confidential communications. The fact that the entire key/password etc wont necessarily be returned, is not a comfort. Because the attacker can simply keep exploiting heartbleed in order to get more data, and ideally all the pieces needed to put the puzzle together.

What I would like to see, is the capture of the random numbers which have been used to create the secret keys. With these, the attacker can generate their own copy of the private key(and perhaps infer some other sensitive details to add to the spoils).

Who is vulnerable?
Heartbeat was introduced with OpenSSL version 1.0.1, and remained unchanged up to version 1.0.1f, so anyone using these versions of OpenSSL are vulnerable. The emergency release of OpenSSL version 1.0.1g on 7th April 2014 fixed the bug, by applying bounds to the heartbeat message. There are a few web services available which you can use to test a website if it is vulnerable(the websites you use, for example). One such service can be found at filippo.io. Java haters will be fascinated to know that the Java standard edition (SE) is unaffected. Microsofts servers are also mostly unaffected, because they tend to use their own implementation of TLS(SChannel). However, other products, such as OpenVPN(which can be used on a microsoft box), do use OpenSSL.

What should you do about it?
The most obvious steps for you to take, would be to change your passwords at any website that has been affected. Indeed, just about every article on heartbleed recommends this. While this is good advice, dont rush out and do it immediately. The reason behind password changes is due to the possibility that your password has already been compromised. But if you change it before the servers OpenSSL has been updated, then your astuteness will not have paid off. If you arent sure that your websites where compromised before the disclosure of heartbleed, you can be sure that there is a much higher chance that they are now. Rather verify that the server has applied the bugfix first(either by testing it with a webservice, or by taking note of any public announcements made by the website).

If you run your own server, you should update your OpenSSL, or recompile it with the -DOPENSSL_NO_HEARTBEATS option, as well as revoking and reissuing all certificates that use SSL/TLS(with new keys, of course. Have fun...), and force client password resets.

One thing which has been bugging me about all the current media coverage to date, is that they are all server centric. I havent seen much mention at all about the possibility that the client is also at risk. Indeed, if a server has been compromised, there is nothing to stop it from using heartbleed against the client. I have found an article about the possibility of a reverse heartbleed attack, where it has been shown that although it is harder to do, the clients(IE. your own devices) are also at risk.

Should we be pointing fingers at anyone about this?
In short, no. Remember that OpenSSL is an open source project. It is the most widely used implementation of SSL/TLS in the world. Not bad for a project run by a core group of four developers, only one of which is considered to be working on it full time. And the pay is next to non-existant. On average, donations to the OpenSSL Software Foundation (OSF) are a meagre $2000 per annum. They do not have the man power to do extensive code reviews, however the larger companies that use the software, do have that capability(and should have reviewed all outsourced code fully anyway, whether the code came from a small group of developers or not).

Some parting thoughts
One question on many peoples minds, is whether or not the NSA has been exploiting heartbleed all along. They claim to have had no prior knowledge of heartbleed, but then they wouldnt admit it if they had known. With their resources and tenacity, I would be surprised if heartbleed was news to them. Other, more paranoid people are saying that since it is known that the US government has worked to weaken encryption standards, they further undermined internet security by not donating any significant amount to the OSF. I think this last is a bit of a stretch. But the real question is, was the bug introduced to the code base deliberately?

Re-issuing certificates is going to be an extremely daunting task to the Certificate Authorities. I doubt they have infrastructure capable of doing all the work in the time frame they have(IE. immediately).

The use of two factor authentication(depending on which kind of 2FA), as well as perfect forward secrecy would have greatly reduced your risk to exposure. Affected servers should be sending out emails explaining the plan of action(change of password etc), but do not trust any link you may find in any of these emails. It could be a phishing attempt(we have all received such emails many times in the past, and this is the perfect opportunity to go phishin').

Lastly, I think it falls on large companies that use OpenSSL to make an effort and donate to the OpenSSL project. The OpenSSL team have done one hell of a job, and have achieved an astounding market share. Perhaps with adequate funding, heartbleed could have been prevented.

Have you got any further questions? Feel free to leave your questions and comments on this page.

Some additional info

Comments (2) Trackbacks (0)
  1. Very interesting, amazingly I understood small portions

Leave a comment

No trackbacks yet.