Heartbleed Bug Vulnerability: Discovery, Impact and Solution

Editor's Notes

1. First a word about the protocol extension that has the vulnerability, then I'll mention the vulnerability itself. The TLS protocol has an extension called 'Heartbeat', defined in RFC6520, designed to allow the use of a keep-alive function for TLS (or DTLS) without performing a costly renegotiation where no session exists (TLS) or could exist (DTLS). Sending HeartbeatRequest messages allows the sender to make sure that it can reach the peer and the peer is alive. Put simply the nature of the request is: 'Hey server, shout the 5 bytes "#CASC" back to me if you're up and listening' A server that is up would send back "#CASC". The Heartbeat extension is actually pretty rarely used and (until recently) rarely mentioned. Now to the Heartbleed vulnerability, and it is the vulnerability we are interested in here. The nature of the Heartbleed vulnerability is that the server will send back more in the reply than it was given in the request in the first place. E.g. 'Hey server, shout the 500 bytes "#CASC" back to me if you're up and listening' A vulnerable server up would send back "#CASC" followed by 495 bytes of values from its internal memory. It is those extra bytes (495 in my example) which can reveal private information which might contain the server certificate's private key, or might contain data from another user's session such as their username and password and credit card number. An attacker gets to try again as often as he likes, each time fetching bytes (up to 64k at a time) from the server.
2. Google and Codenomicon both discovered the vulnerability independantly at around the same time. Google were actually first (April 1st), but Codenomicon (April 3rd) (http://www.codenomicon.com) gave the name Heartbleed (and the logo) to the vulnerability as an aide to promote public awareness. It worked. The idea of a 'branded' exploit is not entirely new, but in this case getting the name and a logo early helped make the vulnerability addressable as a topic of discussion by a wide audience, even those who understood little about its technical nature. The vulnerability was picked up by news agencies across the world - and not just the technical press - and at its peak there was a tweet every second on the subject from regular users - not just from nerds like us. That's great because, as you will hear later, many internet users will be affected directly or indirectly by this vulnerability.
3. RFC6520 specifies the Heartbeat message to have separate length and payload fields. This is not unusual in such protocols. The nature of the bug is that the implementation does not check, when it receives a Heartbeat message, that the length of data it is to return is the same as the length of the data that was supplied to it in the first place (i.e. 500 <> length("Hello")). It accepts the (short) inbound message ("Hello"), and then replies with (e.g.) 500 bytes inadvertently revealing some of its internal state. It is an error in the implementation, not the protocol definition.
4. We don't know! There were some suggestions that examples of logs existed showing misuse of the heartbleed vulnerability, but none of these seem to have been substantiated. It seems likely there is currently no evidence that the NSA (or any other US or foreign agency) created this vulnerability. Human error seems the most likely explanation for it. Nonetheless, while we don't think Heartbleed has been exploited before the time it was announced publicly to be safe we are acting as if it may have been exploited and that leads us to some of the recommendations we will be presenting later in this hangout.
5. Also HP System Management software, procurement software, Canada Revenue Agency (CRA) , VMware series of Horizon products, emulators and cloud computing suites, Western Digital My Cloud product family firmware, HP's BladeSystems, IBM's AIX servers, Dell's appliances and networking equipment Nearly all browsers do not use OpenSSL (exception is Chrome on Android 4.1.1) You’re not secure if you’re using older versions of OpenSSL!
6. http://www.digitaltrends.com/computing/heres-a-list-of-websites-allegedly-affected-by-the-heartbleed- bug/#!FMHrK
7. http://www.forbes.com/sites/bobegan/2014/04/11/a-billion-smartphones-users-may-be-affected-by-the-heartbleed-security-flaw/: Trend Micro scanned a mere 390,000 of some one million apps in Google Play. They found about 1,300 apps connected to vulnerable servers, including 15 bank-related apps, 39 online payment-related and 10 online shopping-related. http://gadgets.ndtv.com/internet/news/android-411-devices-vulnerable-to-heartbleed-bug-says-google-508262 http://thehackernews.com/2014/04/billions-of-smartphone-users-affected_13.html http://www.fiercemobileit.com/story/htc-tops-list-manufacturers-heartbleed-vulnerable-smartphones-says-lookout/2014-04-21 http://btsc.webapps.blackberry.com/btsc/viewdocument.do?noCount=true&externalId=KB35882&sliceId=1&cmd=displayKC&docType=kc&ViewedDocsListHelper=com.kanisa.apps.common.BaseViewedDocsListHelperImpl
8. May not introduce new vulnerability if CDN is just serving static content, but cookies might be sent to CDN
9. http://www.zdnet.com/cisco-juniper-products-affected-by-heartbleed-7000028312/ http://support.f5.com/kb/en-us/solutions/public/15000/100/sol15159.html http://kb.juniper.net/InfoCenter/index?page=content&id=JSA10623
10. https://ics-cert.us-cert.gov/
11. If you are running a web sever, then you need to: Inform, fix, rekey, reissue, revoke, re-inform Inform users of your status. Are you broken, are you shut down, when you will be fixed. Communication is key. Fix the OpenSSL problem. It is likely that your vendor will have a patch. Follow your vendor and install the patch. Rekey your web server. Your private key may have been compromised, so you need a new private key with a corresponding public key. Reissue and install a new certificate with the new public key. Revoke old certificate to mitigate any compromise. Re-inform your users of your status and request passwords be change You may also want to consider mitigating future bugs/attacks by implementing: Perfect Forward Secrecy, Second–factor authentication, end-to-end encryption, or just hardening your server
12. If you are a client As the Heartbleed bug can be used on both a server and a client, does your client software need an update? You shouldn’t have any problems with a browser, but an application may be using OpenSSL As such, check the status or updates to software that you use for secure communications If your web service has told you that they are vulnerable, stop using their service until it is fixed. Once your web service is fixed, change your password If you want to know if a service is vulnerable use a checker to find out; such as the SSLchecker at our CASC Or you can also install the Netcraft app to your browser which will check for you.
13. Since fixing Heartbleed will mean that certificates will be revoked, you want to ensure your browser is checking certificate status. In Windows use Internet Options, select the Advanced tab, and select “Check for server certificate revocation” under Security. There are some debates as to whether users should waste time checking certificate status. We recommend checking status as we don’t believe that you should let Perfect be the enemy of Good.