Encrypted connections established by at least 949 of the top 1 million websites are leaking potentially sensitive data because of a recently discovered software vulnerability in appliances that stabilize and secure Internet traffic, a security researcher said Thursday.
The bug resides in a wide range of firewalls and load balancers marketed under the F5 BIG-IP name. By sending specially crafted packets to vulnerable sites, an attacker can obtain small chunks of data residing in the memory of connected Web servers. The risk is that by stringing together enough requests, an attacker could obtain cryptographic keys or other secrets used to secure HTTPS sessions end users have established with the sites, security researcher Filippo Valsorda told Ars. He didn’t identify the sites that tested positive in his scans, but results returned by a publicly available tool included with his vulnerability disclosure included the following:
- www.adnxs.com
- www.aktuality.sk
- www.ancestry.com
- www.ancestry.co.uk
- www.blesk.cz
- www.clarin.com
- www.findagrave.com
- http://ift.tt/ua88fu
- http://ift.tt/NYCAW0
- http://ift.tt/nEUVCj
- http://ift.tt/PUci8d
- http://ift.tt/LxUxKh
- http://ift.tt/HuxBab
- www.netteller.com
- www.paychex.com
The threat stems from a vulnerability in F5 code that implements a transport layer security feature known as session tickets. Session tickets can speed up encrypted transactions by allowing previously established HTTPS connections without a key having to be renegotiated all over again. Sites that use the vulnerable F5 appliances and have session tickets enabled are vulnerable.
It’s not yet clear precisely what kind of data can be extracted by exploiting the bug. Valsorda, who is a cryptography engineer for content delivery network Cloudflare, said he discovered the flaw by chance as he and a colleague helped troubleshoot error messages received by customer using an F5 load balancer (Valsorda has more details here). So far, Valsorda has observed the bug returning other users’ session IDs, which by themselves isn’t particularly sensitive.
Remember Heartbleed?
Although he has deliberately not attempted to do so, he said he wouldn’t be surprised if the flaw exposed the same types of sensitive information that were exposed by Heartbleed, an extremely high-severity bug in the OpenSSL cryptographic library that came to light in 2014. As a Cloudflare community challenge quickly demonstrated, Heartbleed could be exploited to reveal the secret cryptographic key attackers needed to impersonate a vulnerable website.
“I didn’t want to risk obtaining key material of a third party, and anyway low-level memory analysis is not my expertise,” he told Ars. “The Cloudflare Heartbleed challenge taught us that optimistic assumptions can prove wrong under better scrutiny, so both F5 and I just assumed all memory could be potentially compromised since allocation patterns are undefined.”
The bug is technically known as a buffer overread. It’s the result of F5 developers hardcoding a value of 32 for the length of a Session ID and not accounting for the possibility of receiving shorter lengths. The failure “suggests that F5 software is written in a language that lacks memory safety (possibly C, like OpenSSL and a lot of Internet software today),” Valsorda wrote in an e-mail. “This vulnerability couldn’t have happened in a Go or a Rust codebase. Switching is much easier said than done, but this underscores how important it is.”
F5 has issued a fix for the vulnerability, which is indexed as CVE-2016-9244 and has been dubbed Ticketbleed. Vulnerable sites can also workaround the bug by turning off session-ticket capabilities.
Discussions of the bug on social media are rife with comparisons to Heartbleed, and there are some clear similarities. For instance, they both stem from a vulnerability in widely used TLS implementation that undermine the security of encrypted connections. Both also leak random uninitialized memory, are the result of mistakes made in programming languages that provide no memory safety, and are exploitable using simple code.
But there are also some key differences. For one, the F5 implementation is proprietary and not as widely used as the open-source OpenSSL package. Another is that Ticketbleed exposes much smaller chunks of memory, a trait that requires more effort to exploit. In short, Ticketbleed is no Heartbleed, but it’s still worth patching immediately.
from Ars Technica http://ift.tt/2k80tmV
via IFTTT