Clark Morris wrote, in part:
>Another thing that has always baffled is the idea that even if I have
>a strong password that is NOT written down, I still should change it
>once a month.  If the site I am logging into enforces good management
>by locking the account after say 5 attempts in 15 minutes thus
>allowing no more that 16 attempts an hour or 140544 attempts a year,
>how is not changing my password going to make that much of a
>difference since at 1,404,544 attempts in 10 years that is still a
>small fraction of the 656 billion possibilities with a 8 character
>password assuming ONLY 30 characters in a character set?

I share this reservation to some extent. I believe that for most uses, there 
are (at least) two types of systems:
1) Those we as end-users take seriously
2) Those we as end-users don't take seriously

By that, I mean that class 1 is our bank, work (hopefully!), healthcare 
provider -- systems like that; class 2 is our Outback Curbside Takeway login, 
that online store where you had to register to buy something and will likely 
never visit again, et al.

Now, assuming that people follow such a scheme (and I know many don't--which 
puts all sites into class 2), then I'd argue that the following logic applies.

With class 1, I see no good reason to force frequent changes. Occasional 
changes -- maybe; I'm not entirely convinced but not entirely UNconvinced, 
either. There's some validity to an argument that the longer a password gets 
used, the more likely it is to be accidentally revealed by shoulder-surfing or 
other means, but that seems pretty thin.

For class 2, the argument is that people reuse passwords. They have to: unless 
they use a password manager (which adds its own risk), then who's going to 
remember the login for JoesHamsterShack.com three years from now? No, instead 
we use our "usual" password, adding "1" (or "9", or some other well-remembered 
digit) when it says "Must contain a digit", and some equally consistent piece 
of punctuation when we need that. Similarly, we use the same userid, adding 
digits if needed.

That way we can send ourselves a note saying "JoesHamsterShack.com login: 
usual/usual123", and three years from now, we can look at that note and with 
reasonable certainty say "Ah, right, that means the userid is <x> and the 
password is <y>".

So far so good. But the problem is that Joe sells hamsters: he isn't an IT 
security guy. So when he gets hacked, and it turns out that his user database 
is plaintext--not even hashed passwords--then there's a risk that someone who 
gets a copy of that database will try our usual/usual123 combination on 
SamsSnakeHouse.com and get in (yes, this assumes you have an eclectic pet 
collection!).

How big is that risk? Not clear, in large part because we don't generally find 
out how an ID gets hacked; add in the fact that spoofed email still happens, so 
people often THINK they've been hacked when they haven't really, which muddies 
the waters further.

Yahoo! mail, for instance, is well-known to be "porous", in that we get tons of 
spam from correspondents who use Yahoo! mail, correctly targeted at our email 
(and with other correspondents of theirs on the To: line), which makes us think 
that "their ID must have been hacked--otherwise how would it have known to send 
mail to folks in their address book?" That's a logical conclusion, but it 
happens so often that either Yahoo! is wildly incompetent (possible, but 
unlikely) or there's some OTHER way that folks get ahold of Yahoo! address 
books, without actually getting at Yahoo! email accounts. We just don't know.

So...changing class 2 passwords periodically isn't the world's worst idea. 
Painful? Yes. Every month? Be serious--that's always struck me as stupid: it 
*guarantees* that I'm either going to write them down or use a pattern 
(hamster1, hamster2, hamster3...).

This is all made more complicated when we use multiple, random machines and/or 
share accounts with our spouses: if I kept changing our Amazon password every 
month to one of those recommended random strings of bytes, my wife would be 
irritated (or I'd have to email her the password, which would be insecure). And 
I'd never manage to get logged on anywhere but my home machine!

ObAnecdote: My dad was fluent in many languages. When he taught at the 
University where I worked, he sent me a list of the 12 months in one of those 
languages, and told me that he used one of those as his password; not always in 
sync with the current month, but pretty close. So if I needed to get on one of 
his IDs today (were he alive and I still working there), I'd have tried "may" 
in that language; if it didn't work, "april" or "march" would have. Worked well 
and seemed fairly secure.

As Heather Adkins of Google said (and Heartbleed reinforced), "Passwords are 
dead". The problem is that we don't have a good, universal replacement (yet?).

...phsiii

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to