December 31st, 2008
http://cryptogon.com/?p=5940
“…no tinfoiler in his right mind would bet his life on SSL.”
—The Ugly
Truth About Online Anonymity
* Chortling sounds *
Via: CNET:
BERLIN–A key piece of Internet technology that banks,
e-commerce sites, and financial institutions rely on to keep
transactions safe suffers from a serious security vulnerability, an
international team of researchers announced on Tuesday.
They demonstrated how to forge security certificates used by
secure Web sites, a process that would allow a sufficiently
sophisticated criminal to fool the built-in verification methods
used by all modern Web browsers–without the user being alerted that
anything was amiss.
The problem is unlikely to affect most Internet users in the
near future because taking advantage of the vulnerability requires
discovering some techniques that are not expected to be made public
as well as overcoming engineering hurdles: performing the initial
digital forgery consumed approximately two weeks of computing time
on a cluster of 200 PlayStation 3 consoles. In addition, a criminal
needs to find a way to reroute traffic from a legitimate Web site
to his own, perhaps through techniques that have become well-known
in the last few years.
Yet if one group can do it today, others eventually will.
“We have a proof-of-concept that allows us to impersonate any
supposedly secure Web site on the Internet,” said David Molnar, a
doctoral student in computer science at the University of
California at Berkeley.
Molnar and six other researchers presented their findings
during an afternoon session of the Chaos Computer Club’s annual
conference here on Tuesday. Other team members include Jacob
Appelbaum and Alexander Sotirov.
Their work has focused on finding vulnerabilities in a
technology known as Secure Sockets Layer, or SSL, which was
designed to provide Internet users with two guarantees: first, that
the Web site they’re connecting to isn’t being spoofed, and second,
that the connection is encrypted and is proof against
eavesdropping. SSL is used whenever a user navigates to an address
beginning with “https://”. SSL certificates essentially stand for
the claim that, for instance, etrade.com actually belongs to
E-Trade Inc., and is not being operated by a thief hoping to steal
account passwords.
Most browsers indicate that SSL is active by displaying a
small padlock icon. An attack using a forged authentication
certificate–which is what the researchers say they have done–is
insidious because the browser can’t detect it and the padlock icon
would still appear.
Unlike most security issues, this problem cannot be fixed
with a simple software update. “The bug is not in anyone’s
software,” Sotirov said. “It’s not the browser that’s at fault. The
browser does exactly what it’s supposed to do… The problem is that
what it’s supposed to do is wrong.”
The attack exploits a mathematical vulnerability in the MD5
algorithm, one of the standard cryptographic functions used to
check that SSL certificates (and thus the corresponding Web sites)
are valid. This function has been publicly known to be weak since
2004, but until now no one had figured out how to turn this
theoretical weakness into a practical attack.
An SSL certificate is a small file that ties a real-world
corporate identity to a Web site address and a corresponding public
encryption key. This is presented to a private certificate
authority firm, which is supposed to verify the link between
identity and domain name and then cryptographically “sign” the
certificate to vouch for it.
The problem arises when someone else is able to forge the
same signature.
VeriSign, which operates the largest certificate authority
in the world, learned of the vulnerability early on Tuesday and
acted quickly to close the hole in its certificates, according to
Tim Callan, vice president of product marketing at the
company.
“We went into our systems and removed the MD5 algorith and
replaced it with SHA-1 (Secure Hashing Algorith),” he said. “You
can not get an SSL certificate from VeriSign now that is subject to
this attack.” More information from VeriSign is available on
Callan’s SSL blog.
VeriSign was in the process of phasing out MD5 before the
issue came up and is now on track to have it entirely out of
commission in January, Callan said. “On balance, public key
infrastructure works extraordinarily well,” he said when asked if
the vulnerability illustrated a need to change the trust
model.
Microsoft, while noting that the issue wasn’t a
vulnerability with one of its products, tried to downplay the
threat to users in a security advisory Monday.
“This new disclosure does not increase risk to customers
significantly, as the researchers have not published the
cryptographic background to the attack, and the attack is not
repeatable without this information,” the advisory said.
A 1991-era protocol, but modern problems
When MIT professor Ron Rivest developed MD5 in 1991, it was
considered sufficiently secure. But starting in 1996, a series of
increasingly serious flaws started calling the continued viability
of MD5 into question.
As CNET News reported in 2004, flaws discovered at that time
“could eventually make it easier for intruders to insert
undetectable back doors into computer code or to forge an
electronic signature–unless a different, more secure algorithm is
used.” Then, in 2007, Arjen Lenstra of Bell Laboratories
Switzerland, with Marc Stevens and Benne de Weger of TU Eindhoven,
demonstrated a technique to construct two new certificates with
different content but the same fingerprint.
Although security researchers had been worrying, and
recommending that other alternatives be considered, nobody had yet
demonstrated how to exploit this theoretical flaw in a practical
attack.
Molnar, Appelbaum, and Sotirov joined forces with the
European MD5 research team in mid-2008, along with Swiss
cryptographer Dag Arne Osvik. They realized that the
co-construction technique could be used to simultaneously generate
one normal SSL certificate and one forged certificate, which could
be used to sign and vouch for any other. They purchased a signature
for the legitimate certificate from an established company that was
still using MD5 for signing, and then applied the legitimate
signature to the forged certificate. Because the legitimate and
forged certificates had the same MD5 value, the legitimate
signature also marked the forged one as acceptable.
The process amounted to transferring a photograph from a
real ID to a fake by carefully matching the holographic security
markers.
The rogue certificate can then be used to sign any other
certificate of the attacker’s choosing–such as one which assures
Web browsers that a malicious phishing site is actually the
legitimate etrade.com or bankofamerica.com.
After three unsuccessful attempts, each of which required
approximately three days of compute time on a cluster of 200
PlayStation 3s, the researchers obtained a forged certificate
authority in early November, at which time they notified browser
developers and certificate authorities, or CAs, about the security
flaw. Molnar estimates that the same processing time could be
purchased from Amazon for about $1,500.
The team decided to disclose the vulnerability at the Berlin
conference in hopes that the news will encourage everyone involved
to fix the problem quickly. “The main message here is to stop
issuing MD5 certificates, now,” said Molnar. He believes that MD5
is so weak it no longer should be used for any applications: “More
secure, freely available alternatives exist.” (In November 2005,
the U.S. government announced plans to find successors to MD5 and
SHA-1, an official federal standard with its own problems. The new
federal standard will be called SHA-3.)
By itself, the MD5-certificate-forging vulnerability
wouldn’t be too worrisome. That’s because it relies on criminals
being able to capture Web traffic to display a fraudulent Web site.
But setting up a fake wireless access point to lure unsuspecting
neighbors or business travelers is trivial, and a program released
earlier this year to attack the domain name system (DNS) provides
another way to direct Internet traffic for malicious
purposes.
While only a few CAs currently sign certificates with MD5,
Appelbaum estimates that 30 percent to 35 percent of all SSL
certificates currently in use have an MD5 signature somewhere in
their authentication chain. “The CAs should contact every customer
that currently uses an MD5-signed certificate and offer a free
replacement.”
In an interview on Tuesday morning, cryptography expert
Bruce Schneier praised the research but downplayed the real-world
consequences of the findings.
“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,” he said.
“This is good work, great cryptography. I love the research,
but this doesn’t matter a whit,” Schneier added. “There are half a
dozen ways to forge certificates and nobody checks them
anyway.”
Paul Kocher, president of Cryptography Research and an
architect of the SSL 3.0 protocol, said the exploit highlights the
need for a new universal hash function “that everyone is
comfortable with.”
“The paper is not a surprise, but at the same time it’s the
crispest demonstration for why it’s necessary to remove this broken
algorithm everywhere it is being used,” he said, before adding
“there are bigger things to worry about, like browser bugs and
operating security bugs.”
The researchers have created a Web site signed with a forged
certificate which can be viewed here. The forged certificate was
backdated so that it could not be used maliciously even if stolen
from researchers, so you have to reset your system clock to August
2004 to view it.
Even though their work may be controversial, the researchers
view their efforts as fundamental to creating a more secure
Internet. “I don’t want to be hit by this type of attack either,”
Sotirov said. “I use the Internet too.”