W3C home > Mailing lists > Public > public-wsc-wg@w3.org > April 2007

Re: Shared Public Knowledge

From: Mary Ellen Zurko <Mary_Ellen_Zurko@notesdev.ibm.com>
Date: Tue, 24 Apr 2007 08:23:08 -0400
Cc: public-wsc-wg@w3.org
Message-ID: <OF623E1CC3.05F377DC-ON852572C7.0043F3D1-852572C7.00440A64@LocalDomain>
To: "Chuck Wade <Chuck" <Chuck@Interisle.net>
Side comment - on your preventing MitM comment - anyone know of 
technologies or other work out there? It would be worth pointing to in our 
Futures section of the wiki. 


Mary Ellen Zurko, STSM, IBM Lotus CTO Office       (t/l 333-6389)
Lotus/WPLC Security Strategy and Patent Innovation Architect

Chuck Wade <Chuck@Interisle.net> 
Sent by: public-wsc-wg-request@w3.org
04/14/2007 11:44 AM

Mary Ellen Zurko <Mary_Ellen_Zurko@notesdev.ibm.com>
Re: Shared Public Knowledge


The "last login" message is useful as both a means for confirming site 
authenticity to the user as well as letting a user detect unauthorized use 
of their account. This convention goes way back to some of the earliest 
time-sharing systems, and was originally motivated by the desire to detect 
misuse of expensive computing resources. Later, it became an indirect way 
to confirm site authenticity, since an impostor generally has no way of 
knowing the last time someone logged into an account. 

However, this is again only effective if users pay attention to the "last 
login" message. Bank of America and many other financial sites do use this 
technique, though it is independent of other authentication measures, such 
as SiteKey. The key point, though, is that real world authentication 
depends on lots of contextual clues that observant participants can use 
throughout the course of a dialogue. Many financial institutions now 
monitor many aspects of the communications channel with a user along with 
user behavior to determine if there are reasons to suspect improper 
access. However, users do not have similar resources at their disposal to 
analyze the sites they communicate with.

Of course, most users probably ignore contextual information, such as a 
"last login" message. At the same time, this is a no-cost option, and it 
does get used in some situations that are important to financial 
institutions. Since a high percentage of the fraud that takes place today 
is perpetrated by "friends and family," the "last login" clue will be used 
by someone who suspects that a friend or family member is accessing their 
account without permission. In other words, it's a clue people use when 
they have reason to suspect a problem.

Similarly, if someone suspects that they might have just logged into a 
bogus Web site, they might look at the "last login" message to confirm 
their suspicion, or to restore confidence so that they can proceed. I 
doubt if anyone has studied the effectiveness of this measure, but it 
probably helps in a small percentage of cases. Again, it's a no-cost 
option, so it delivers a high benefit/cost ratio, even for small benefits.

It is also worth noting that legitimate sites that have a prior 
relationship with a user tend to present a lot of context to the user that 
helps them establish confidence in the site. If I access my bank accounts 
online, I will quickly see a bunch of indicators that could only have come 
from my bank, such as balances and the arrangement of account information. 
Similarly, good practices among online retailers involve demonstrating to 
the customer that they have been recognized, and presenting lots of 
information that is really unique to that customer (e.g., recently viewed 
items, order status). 

Now, if you'll forgive the tangential observation, I think it is helpful 
to recognize that human beings are complex entities that survive through a 
variety of trained responses to an extremely broad array of contextual 
clues. Sure, simple interfaces are best, but finding one set of indicators 
that will work for all users in all transactional settings is a very 
difficult problem. An alternative approach is to provide a broader set of 
indicators more tightly integrated into the context of the interaction as 
a way to trigger both positive and negative confidence assessments on the 
part of real human users and their robust pattern matching abilities. We 
recognize who we're dealing with when interacting with other people 
through an amazingly complex set of pattern matching operations that 
process enormous amounts of input data through multiple sensory channels. 
The problem with the cyber world is that digital patterns really are all 
the same. However, by introducing contextual clues that vary from 
circumstance to circumstance, we may re-enable human abilities to detect 
patterns that are either consistent or inconsistent with our expectations. 
Perhaps this is an alternative approach to user interface design that 
could lead to more effective user determination of site authenticity.

The real challenges, though, come back to Man-in-the-Middle attacks and 
purloining of userids and passwords by bogus sites. Where we most need 
purely technical counter measures is in detecting and preventing MitM 
attacks. However, if a user can quickly detect that something is wrong 
when they land on a bogus site, then there is both greater likelihood that 
corrective actions can be taken in a timely manner, and that the bogus 
site will be blocked or taken off line. Ultimately, adequate security and 
safety require on a "defense in depth" strategy that employs many 
complementary measures. We just need to find ways to better engage the 
survival skills that we human beings have evolved over eons as part of the 

   Chuck Wade, Principal
   Interisle Consulting Group
   +1  508 435-3050  Office
   +1  508 277-6439  Mobile

Mary Ellen Zurko wrote: 

I disagree, and if it makes sense as a site to user antipattern (and I 
sense the jury still out on that), if there is concensus, we can say 
something appropriate about what, if anything, should be implied for the 
other direction (and the going in position from me would be, nothing 
should be implied for the other direction). 

What things other than SiteKey use information (secret, public, or shared 
public) to (attempt to) authenticate the site to the use? Anyone have more 
examples? Thanks Chuck for the Sitekey one. And Chuck, is the last login 
time _really_ meant to authenticate the site to the user? I thought it was 
to give the user a hint if the account had been unknowingly used by 
someone else. 


Mary Ellen Zurko, STSM, IBM Lotus CTO Office       (t/l 333-6389)
Lotus/WPLC Security Strategy and Patent Innovation Architect

04/12/2007 07:34 PM

RE: Shared Public Knowledge

Thanks for this clarification.  But my concern is if W3C declares SPK 
based site-to-user authentication to be an anti pattern, that certainly 
implies it should never be used in the other direction either.

From: Mary Ellen Zurko [mailto:Mary_Ellen_Zurko@notesdev.ibm.com] 
Sent: Thursday, April 12, 2007 3:17 PM
To: McCormick, Mike
Cc: public-wsc-wg@w3.org
Subject: Re: Shared Public Knowledge

I would like to do a rewind on this thread. Everyone who participated, go 
back to the proposed recommendation that we discussed:


It's about authenticating the server to the user (since that's one of our 
primary goals). Not the user to the server. 

So I will assume all discussion of the latter was interesting and 
informative (it was for me), but not about the actual proposal being 
discussed. Maybe that's because the proposal is about something nobody 
does or wants to do. That would make it nice and safe for our 
recommendations :-).


Mary Ellen Zurko, STSM, IBM Lotus CTO Office       (t/l 333-6389)
Lotus/WPLC Security Strategy and Patent Innovation Architect

Sent by: public-wsc-wg-request@w3.org 
04/11/2007 07:47 PM


Shared Public Knowledge

I had to drop off the line for a few minutes at the top of the hour during 
this morning's meeting.  Regrettably that moment came during the Lightning 
Discussions just as Chuck Wade was responding to MEZ's presentation on 
Shared Public Knowledge (SPK).  By the time I rejoined to discussion had 
moved on to the next topic. 
What I would have said given the opportunity is that Chuck is 100% right. 
In our industry this battle has been fought many times and I see little 
good coming from taking a hard line against all online use of SPK. 
Many US companies rely on services provided by the likes of Choicepoint & 
Acxiom to perform Knowledge Based Authentication (KBA) or Out of Wallet 
Authentication (OOWA) of consumers in certain situations, especially in 
cases where no prior business relationship exists between the FI and said 
These KBA systems typically ask a series of randomly chosen multiple 
choice questions designed to score a user's knowledge of semi-private 
information about himself or herself.  Examples might include "What model 
car do you drive"? or "What¡¦s the amount of your monthly mortgage 
payment?".  A determined criminal could undeniably obtain this information 
from public sources, perhaps even use it to impersonate others, but that 
doesn't mean there is no legitimate use case for KBA. 
A blanket prohibition against KBA is unnecessary and would never be 
accepted.  Asking the user enough SPK based questions is not an 
unreasonable authentication technique as long as the associated risk is 
low, or when SPK is only being used to supplement some other credential 
for extra assurance. 
The much maligned Mother's Maiden Name is an example of weak KBA ¡K but 
much stronger ones are possible using the enormous databases of personal 
data that are available from brokers today.  So I think the SPK 
"anti-pattern" would benefit from being softened a bit to acknowledge 
there's a place for it under certain conditions.
Thanks, Mike 
Michael McCormick,CISSP
Lead Architect, Information Security Technology 
Wells Fargo Bank 
255 Second Avenue South 
MAC N9301-01J 
Minneapolis MN 55479 
(ƒn      612-667-9227 (desk)             7      612-667-7037 (fax) 
(       612-590-1437 (cell)             J      
michael.mccormick@wellsfargo.com (AIM) 
2       612-621-1318 (pager)            *      
This message may contain confidential and/or privileged information.  If 
you are not the addressee or authorized to receive this for the addressee, 
you must not use, copy, disclose, or take any action based on this message 
or any information herein.  If you have received this message in error, 
please advise the sender immediately by reply e-mail and delete this 
message.  Thank you for your cooperation. 

Received on Tuesday, 24 April 2007 13:38:28 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:36:44 UTC