Slashdot Mirror
Analyzing 20,000 MySpace Passwords
Rub3X writes "Author found 20 thousand MySpace passwords on a phishing site and did some tests on them. They were tested for strength, length and a number of other things. Also tested was the most popular password, and the most popular email service used when registering for myspace."
Say, 10% of passwords contained on a site was obtained using a dictionary attack. Then perform analysis on these password. Conclusion that basing on statistically significant number of passwords (10%, >10000) almost 100% of passwords on the site are vulnerable to dictionary attack is simply wrong - the sample was biased.
Similar about phishing-originated passwords. Phishing is a result of bad practices on user side, and usually clicking attachments in spam, using insecure browser and no antivirus is connected with using poor quality passwords. The results WILL show worse quality of user passwords than real simply because the passwords originate from subset of users who know less of security in general (and as result, got hacked.)
Anagram("United States of America") == "Dine out, taste a Mac, fries"
It would be interesting to see how many of the names in that list use the same password for MySpace account as they do in their email account.
There was an MIT study claiming that the strength of passwords was affected by length alone. Because of brute force cracking, the longer the password, the longer it took to break. Consider the three character password where I allowed only numbers, and upper and lower case letters. Each position in the password would have 10 + 26 + 26 = 62 possibilities. A three letter password would have 62 * 62 * 62 combinations. Now, if I required "strength" by requiring the use of a letter, and both upper and lower case, I now have only 10 * 26 * 26 combinations. Requiring "strength" always reduces the set of possible combinations for the password.
Twenty-two years on, here's my obvous password detector. This is C source code I wrote in 1984. This simple piece of code will prevent the use of passwords that are English words, by requiring that the password have at least two sequences of three letters not found in the dictionary. The "dictionary" is compressed down to a big table of hex constants; it's a 27x27x27 array of bool, with a 1 for each triplet found in the UNIX dictionary. So the code is simple, self-contained, and does no I/O.
Put this in your password-change program and dictionary attacks stop working.
The code is a bit dated; this is original K&R C, not ANSI C.
I should do a Javascript version and give that out. The code is so small that it could easily be executed on user-side password pages.
He came up with a rating scheme from 1 to 4, where 4 is the "best" password. And he says "I consider strength two fine for a myspace account." Very good point: Not all websites need the same level of password strength.
My personal pet peeve is websites that probably only require a 2 or 3 (on his scale) but demand strength 99. For example, forum sites that reject passwords that my bank would consider good enough.
My plea to anyone reading this who develops websites: The strength of the password only has to match the importance of the information that it's protecting.
Thus endeth my rant.