Forwarded from: security curmudgeon <jericho (at) attrition.org>
ISN posted the article as well, cross-posting my reply.
: Security flaws plague majority of e-banking sites
: http://www.finextra.com/fullstory.asp?id=18764
:
: Over 75% of banking Web sites contain fundamental design flaws that
: could put customers at risk from cyber thieves, according to a study
: (of 214 bank web sites)conducted by researchers at the University of
: Michigan.
Unfortunately, all I can find are articles mentioning the study. It
still isn't available on Atul Prakash's home page [1]. Since all we have
to go on right now are sound bytes and brief summaries, it is very easy
to tear large holes in the results. I encourage Prakash and his team to
make the original research more readily available.
: The flaws are not bugs that can be easily fixed with a patch, but are
: systemic, stemming from the flow and layout of the sites.
The flaws are often very easy to fix, and do not require much work from
the bank.
: 47% placed secure login boxes on insecure pages.
While a bad practice, this doesn't translate to "attackers can get
access to customer information" necessarily. "He says this allows
hackers to re-route data entered in the boxes or create a spoof page to
harvest information." First, to re-route data entered in the boxes
relies on something more than a mixed HTTP/S page. Exploiting
cross-frame scripting in some browsers would be a good idea, but that
can be blocked regardless of the page being served over SSL. Second, bad
guys can spoof pages regardless of the presence of SSL, yet Prakash
suggests otherwise.
"Prakash says in a wireless situation, it's possible to conduct this
man-in-the-middle attack without changing the bank URL for the user, so
even a vigilant customer could fall victim."
Certainly a risk, but the amount of customers accessing their bank over
unsecured wireless are probably very minimal and changes the
requirements of exploitation considerably.
: 55% put contact information and security advice on insecure pages.
Again, having a static /contact.html on the legitimate domain, not
served over SSL is not a vulnerability, and does not lead to customers
being at risk from "hackers getting access to customer information". The
summary and introduction to the article is poorly worded and misleading.
: Some banks use social security numbers or e-mail addresses as user
: IDs.
This is definitely a bad practice and commonly seen, but this is half of
the information needed to authenticate. Brute forcing a list of login
IDs is time consuming, brute forcing valid passwords for them on top of
that is very time consuming. There are certainly controls that can be
put in place to make harvesting attacks more costly, regardless of the
login name scheme.
: 28% don't state a policy on passwords, or allow weak passwords.
Yes, they should state their policy, but how many of the 28% don't state
the policy yet enforce a relatively strong one? This number is a poor
metric.
I have a hard time believing that Prakash and his team got permission to
test 214 bank web sites. If they did, it was still done without
authentication based on the results in the article. The few results they
do have are not near the risk implied by the summary wording or have
caveats on exploitation. None of them are real eye-openers as each one
would likely result in the compromise of a handful of accounts. While
certainly bad, that is insignificant compared to an SQL injection or
privilege escalation attack that allowed cross-user information
disclosure (or manipulation).
All said and done, this research is quite limp so far.
- jericho
Received on Fri Jul 25 07:30:40 2008