MySpace password exploit: Crunching the numbers (and letters)

I didn't intend to discuss passwords again this week, but a unique opportunity presented itself. A major phishing attack occurred at MySpace, and I got over 34,000 real passwords to analyze for character frequency.

Free IT resource Virtualization Insights from Top Experts - Learn how virtualization gets real! Sponsored by Dell Free IT resource TechNet: More ways to know it, share it, and keep it running. Sponsored by Microsoft

The phishing attack occurred because hackers were able to use MySpace’s HTML home page abilities to craft a malicious overlay page. That meant when a MySpace user thought they were logging in to a friend’s MySpace home page or profile, they were often sending their log-on names and passwords to hackers, who collected them on other compromised Web servers. It is estimated that the hackers collected more than 100,000 log-on names and password combinations before the phishing attack was noticed.

The hackers erred in making the collected passwords available for anyone to see and download. There were at least five different collection points. I picked up the password files from two of the locations; they came in at over 2GB.

After removing tons of garbage (some from coding errors, some from helpful people trying to crash the hacker’s collection files), I came up with more than 34,000 different log-on account/password combinations. Being able to collect and analyze such a large number of passwords from a wide range of users doesn’t usually happen when you’re on the white-hat side of things.

I collected the passwords into MS Access and MS Excel databases and then analyzed them for word and character distribution. Here are some of my findings from my initial queries:

*As expected, English vowels are by far the most frequent occurring password symbols (E, 48 percent; A, 46 percent; I, 34 percent; O, 33 percent). Other high-ranking letters included R (35 percent), S (32 percent), N (31 percent), L (28 percent), T (25 percent), C (21 percent), and M (21 percent).

*The letters, B, D, G, H, P, U, and Y appeared in 10 to 20 percent of the passwords.

*As expected, the letters Q (1 percent), X (3 percent), and Z (3 percent) were not popular.

*Numbers were used in well over half the passwords. The number 1 appeared 45 percent of the time, followed by the numbers 2 (22 percent), 0 (16 percent), and 3 (15 percent). Numbers 4 through 9 appeared roughly 9 to 11 percent of the time.

*As I’ve written many times, including in my last column, numbers are most often placed at the end of the password when used. For example, when the number 1 appeared, it only showed up 7 percent of the time as the first character, and only 15 percent of the time as one of the first four characters in the password.

*MySpace accounts don’t require complex passwords, so capital letters and other keyboard symbols -- such as ~, !, &, @, #, and so on -- were not present most of the time. The exclamation point was the most commonly used non-alphanumeric character at almost 3 percent, followed by the period symbol at 1.6 percent.

*Almost 1 percent of users had the word "password" as, or as part of, their password. Not real clever.

*Words, colors, years, names, sports, hobbies, and music groups were very popular. FYI, your girlfriend or boyfriend’s name isn’t that uncommon in most cases. I, too, luv Brandi, Bob, or Joe.

*The color red was twice as likely to be used in a password as blue. No other colors came close in popularity percentage-wise. I guess "chartreuse" is a relatively safe password choice.

*Other popular words include: angel, baby, boy, girl, big, monkey, me, and the.

*Cuss words were very popular. Boy, there’s a lot of aggression out there.

*I was surprised about how many Christian-sounding -- for example, "Ilovejesus" -- log-on names were associated with the worst cuss words.

*Names of sports -- golf, football, soccer, and so on -- were as popular as professional sports teams and college team nicknames.

*Certain specific letter combinations -- aa, ee, oo, dr, ea, lo, la, and so on -- appeared in a given password about 3 percent of the time.

One last note: The password list contained several e-mail/log-on account names from popular OS and software vendors. Although we can’t be assured that the passwords used on the exploited site were the same as the employee’s company password, I’m sure some are matches.

Remember this and learn from it: An exploited Web site that's completely unrelated to your company could still put your company at risk. Remind all employees not to use their company passwords on noncompany Web sites, if at all.

After going through all the files, I revealed no startling password distribution data. All of it backed up my previously published conjecture and studies (such as Perfect Passwords by Mark Burnett and The Great Password Debates by Dr. Jesper Johansson) by other friends.

And in case you’re wondering, hard-working network and security experts spent many hours notifying the ISPs and affected companies about their compromised users' passwords. Of course, I’m willing to bet that a moderate percentage of those contacted will not change their password because they will think the warning notice from their ISP is a phishing message. So they will delete it without responding or changing their password. That’s the world we live in at the moment.