orbitalflower

When can you use a weak password?

Posted in Opinion on

It’s possible to objectively measure the strength of a password against brute-force attack, but since strong passwords also tend to be difficult to remember it’s worth considering the situations in which a weaker, easier password would be sufficient.

Risk level

There are roughly three situations in which an attacker tries to obtain access to your account or system:

Situation 1: Indirect access

In this situation, someone attempts to log in to your online account, but due to technical limitations and the provider’s security precautions they can only guess a relatively small number of passwords per minute.

Ideally, you should only have to remember the password to your main e-mail account, which should be at a reliable provider like Yahoo or GMail. All other passwords can be stored in your password manager or web browser, since you can reset account passwords as long as you can access your e-mail. Most accounts, then, can have randomly-generated strong passwords that you don’t need to worry about.

A major e-mail provider is likely to limit the number of attempts an attacker can make per second. Google, for example, starts showing CAPTCHAs which prevents automated software from logging in at a high rate. Even websites without this security measure, it’s probably not feasible for an attacker to make as many as a million login attempts per second. One attempt per second is more plausible.

An attacker making one attempt per second could guess every commonly-used English word (about 100,000 possibilities) within a day. Within a year they can make around 224.9 attempts, enough to guess every common word followed by two digits and nearly every combination of two common words. Weak passwords are clearly insufficient even on a secure site.

If able to make 1,000 logins per second, even after ten years the attacker could make only around 238 password guesses, which is just under a 1% chance of guessing a password from a search field of 245 guesses. A moderate password is potentially under threat here, but a stronger one (e.g. xkcd-style four common words) is outside of this range. This is very approximately the upper bounds of what an attacker can do without gaining direct access to a system.

Situation 2: Direct access

An attacker has physical access to some encrypted data or system: they’re within physical range of your Wi-Fi signal, have stolen your encrypted smartphone or PC, copied an encrypted file from cloud storage or downloaded a website’s hashed password list.

They can now make brute force password attempts as fast as their hardware allows, and for as long as Moore’s Law holds up, the power and cost-effectiveness of brute force password cracking systems will increase exponentially over time.

According to records on Wikipedia, in 1997 a group of Beowulf clusters could be built with a top speed of around 25 gigaFLOPS for a million dollars in today’s money. By 2007, a cluster capable of the same could be built for a mere $1,265 - over 800 times more calculation per dollar. By 2013, the same $1,265 could buy 11,265 gigaFLOPS.

In practical terms, oclHashCat gives the following benchmarks for a $5,000 system with eight AMD R9 290X graphics cards capable of approximately 45 teraFLOPS:

Encryption type one second one day one year
WPA/WPA2 1.536m 237 245
Whirlpool 1122.304m 246 254
MD5 92672m 252 261
NTLM 175808m 253 262

Even with WPA2, a weak Wi-Fi password is broken within days. Passwords for Whirlpool-based disk encryption are broken up to 8 characters full alphanumeric within a week and 9 characters within a year. MD5, still used by many websites to hash their passwords, breaks every 10-character alphanumeric and 13 character alphabetic within a year, and in practice even more than this since passwords tend to follow predictable patterns.

And attacks will only grow faster and more capable with time and money. A nation state, large company or criminal group can use million dollar hardware and add about 8 bits. Each year, Moore’s Law adds about 1 bit, and spending a year instead of a day adds about 8.5 bits. There’s also a 1.6% chance of finding something that should be 6 bits above capability. And, on average, something will be found halfway through the set, effectively squeezing out another 1 bit.

Strong passwords (75 bits and higher) are vital if an attacker can make direct access to a system. Weak passwords are trivially broken, and moderate passwords with sustained effort.

Situation 3: Immediate access

An attacker has gained the password already through some means, or else has acquired the unencrypted data or access to an unencrypted system. Either way, the password strength is now completely irrelevant.

In this case, other protective measures are necessary. Encrypt data where possible, and take steps to safeguard both the software of your device and its physical security. If the attacker can gain physical access to unprotected data, the password is irrelevant.

A corollary of this is that an attacker can steal a password from a site that uses poor security, then go on to use that password successfully at a site where password cracking isn’t an option. Password re-use is therefore an extremely poor idea (if you can only remember one password, use a password manager software).

Conclusion

See also