|
useit.com |
| Search |
Jakob Nielsen's Alertbox, November 26, 2000:
Summary:
A big lie of computer security is that security improves as password complexity increases. In reality, users simply write down difficult passwords, leaving the system vulnerable. Security is better increased by designing for how people actually behave.
Usability advocates and security people have opposite goals that create a fundamental conflict:
In reality, passwords that comply with the above list of "security-enhancing" principles lead to one outcome: Users write down their passwords. Take a walk around any office in the world and you can collect as many passwords as you like by
When you require simple passwords that users can remember, you increase the probability of passwords being kept secret. The same goes for passwords that users choose and that they don't have to change too frequently.
While it's true that such passwords are easier to crack, the vast majority of security breaks come from intruders (or insiders) who expose a human weakness, not those who run code-breaking algorithms.
In the future, security will improve through biological verification mechanisms, such as fingerprint recognition or retina scanning. However, it will take time for this infrastructure to be built (and fingerprint systems won't work for some people). In any case, for now, it is best to avoid imposing awkward log-in procedures and instead simply cookie users on systems with low security needs.
With ecommerce in particular, users should not have to create a userid and password before they can shop. How many sales are lost because shoppers either cannot think up a unique userid or don't understand how to deal with passwords?
Nonetheless, I do recommend letting users enter their email address instead of a userid: It's guaranteed to be unique and it is easy to remember. However, you should always recommend that they choose a new and different password. Also, in usability testing, I've found that some users expect the system to create the password and then e-mail it to them. You should thus make it explicit that they must create their own while registering or you risk having users either get stuck or close their browser before completing their account set up.
Many websites have harsh requirements for password format. I suggest that you relax the rules as much as possible. Obviously, a system for trading millions of dollars must be more secure than one that lets people read the daily news.
If your rules are too strict, many users will not be able to use names and passwords that make sense to them. This increases the likelihood of users forgetting their login information the next time they visit. Forgotten passwords are the cause of countless repeated registrations across the Web: People often have five to ten "accounts" on the same website.
You should place instructions for userids and passwords immediately next to the field label:
Any other placement and many users will not read the instructions.
A key issue here is the definition of "the system." Ideally, the system is the users' total experience and thus they should only have to identify themselves to the computer once. The fact that they are actually browsing multiple websites should not be the users' problem. Indeed, in the future, personal computers will probably become truly personal and serve as users' agents in cyberspace, including taking over responsibility for passwords and identification.
For now, at a minimum, "the system" should include everything under a user's control. This means, for example, that users should have a single log-in for viewing their account and trading, and a single log-in for entering an extranet and checking on the status of an order. Although the back-end might need to enforce certain privileges for certain users, it should do so transparently without the need for additional log-ins.
[Update added 2004: Our research study on intranet usability, testing the intranets across a large number of companies, found that the sign-on process had the second-largest impact on employee productivity of the factors we tested. (Search was #1.) The difference in sign-on usability between intranets in the top 25% and intranets in the bottom 25% amounted to $2.5 million per year for a company with 10,000 employees.]