Re: Interesting Research

[Mark Poepping <poepping () CMU EDU>]

Just my opinion...  Based on the work done locally and the assistance I've provided over the years...

This is an unnecessary risk.  We first said 'no' to a similar data request over 25 years ago.  I don't believe there is 
any reason for anyone to collect and hold passwords, encrypted or not (except the password tools a user might use, or a 
recovery vault with an enterprise key).

Our general approach is to respectfully ask a researcher: what might be the minimal data needed to test the hypothesis? 
 We have found most security researchers to appreciate this approach and come back with a *much better* proposal (or 
not come back).  Several years ago we supported an experiment to test the effective entropy in passwords actually 
selected by local users (driven by a 'global reset') - the same base data that you're talking about, I think - we did 
it all without anyone seeing or ever storing passwords (researchers wrote all the code - we vetted the safety and 
anonymity of the data export, with IRB approval).  I'd start by suggesting they calculate a strength vector with a 
unique but unmappable identifier to support longitudinal analysis (e.g. give them a username if this is a controlled 
test) - that should make the data safe, period.

To two of your specific points:
#2 - there are several ideas: test the strength of 5000 password from users without training and 5000 with training, 
compare results; or test 5000 untrained users, then train them and test again.  This is a little weird anyway, since 
most password systems enforce some rules, and that'll affect applicability of the results.
#3 - usually people will want username as a dangerous indicator of uniqueness (map pre-training to post-training) - you 
can do the same with a non-reversible (and non-public) hash - be careful to avoid replay because it defeats 
non-reversibility.  i.e. If I have your list of usernames and your hash function, I can map all your usernames to the 
hashes - don't need to reverse it.

Finally, be careful of the claim that it's "just an experiment, the passwords aren't real, we just ask them to make one 
up"...  Even if you tell people not to use any password they currently use, there's no assurance of that so the data is 
still a risk.  Basically, collect only what you really need.

Hope it helps.
Mark.


From: The EDUCAUSE Security Community Group Listserv <SECURITY () LISTSERV EDUCAUSE EDU> On Behalf Of King, Ronald A.
Sent: Tuesday, April 9, 2019 2:54 PM
To: SECURITY () LISTSERV EDUCAUSE EDU
Subject: Re: [SECURITY] Interesting Research

Thank you all for responding. This is great info I will be sharing with the researcher, our IRB, our CIO and possibly 
legal.

Some quick replies to questions:

*         The IRB has yet to approve this research.

*         Using a tool to analyze a password to detect strength: the idea of the research is to see if training on 
strong passwords paid off. Such a tool may or may not work as I would think we would need to know the strength before 
and after the training.

*         I am not sure why they need username. I will follow up.

*         I hadn't thought of the privacy impact, but, now that I have, the risk has grown.

o   (I foresee) Then getting consent or warning the student ahead of time would logically lead to a student adapting 
momentarily and invalidating the research.

*         I do not know if the VCR/VPR is involved in this. Our CIO knows and is of the same mindset I am.

I still keep coming back to the password reuse. In my 3 months back here, I have talked with one student who's banking 
info was changed in our student system. Should this DB end up in the wrong hands, it will get worse.

Reading list (Thank you much for these!):
https://dl.acm.org/citation.cfm?id=1242661 (Microsoft password reuse, 2006)
https://www.blaseur.com/papers/chi16-pwperceptions.pdf, "Do Users' Perceptions of Password Security Match Reality?"
http://www.cs.unc.edu/~reiter/papers/2010/CCS.pdf, "The Security of Modern Password Expiration: An Algorithmic 
Framework and Empirical Analysis"
https://www.cs.umd.edu/~jkatz/security/downloads/passwords_revealed-weir.pdf, "Testing Metrics for Password Creation 
Policies by Attacking Large Sets of Revealed Passwords"
https://www.nsf.gov/bfa/dias/policy/human.jsp (NSF Human Use)
https://www.slideshare.net/JISC/password-lifespans-at-ucl-a-training-opportunity, (UCL in the UK sliding scale of 
password strength and renewal time frame)

Thanks again for all the valuable input!
Ron

Ronald King
Chief Information Security Officer

Office of Information Technology
(757) 823-2916 (Office)
raking () nsu edu<mailto:raking () nsu edu>
www.nsu.edu<http://www.nsu.edu/>
@NSUCISO (Twitter)
[NSU_logo_horiz_tag_4c - Smaller]

From: The EDUCAUSE Security Community Group Listserv <SECURITY () LISTSERV EDUCAUSE EDU<mailto:SECURITY () LISTSERV 
EDUCAUSE EDU>> On Behalf Of King, Ronald A.
Sent: Tuesday, April 2, 2019 4:01 PM
To: SECURITY () LISTSERV EDUCAUSE EDU<mailto:SECURITY () LISTSERV EDUCAUSE EDU>
Subject: [SECURITY] Interesting Research

Fellow security pros,

I have an interesting research request come in my inbox today. A researcher wants to setup a portal for students to 
self-register with a username and password. The kicker is passwords will be stored in plain text and collected. The 
premise is to gauge whether students are actually adhering to suggested practices in password design.

My first reaction is "(heck) no," but I realize I may be overreacting. So, I decided to see if anyone has dealt with 
this kind of research and how you handled it.

While I see the value in the research, my security senses tell me students will be using their standard password they 
use for everything. Thus big risk.

Feel free to contact me directly.

Thank you,
Ron

Ronald King
Chief Information Security Officer

Office of Information Technology
(757) 823-2916 (Office)
raking () nsu edu<mailto:raking () nsu edu>
www.nsu.edu<http://www.nsu.edu/>
@NSUCISO (Twitter)
[NSU_logo_horiz_tag_4c - Smaller]