An alternative approach to protection against leaving a webapp without logging out

One of the most controversial requirements on a web application is to deny returning of a user after that user has left the domain of a web application. It is a very common requirement among security conscious customers because it prevents an abuse of a session of a user who continues his browsing in the same browser window without logging out.

Sadly, this requirement goes against basic principles of the web and browsers not only doesn't give you any means to achieve it, it also actively fights against this intention. How should we deal with it? Very common implementation of this requirement is to encapsulate the application into one big html frame. This solution has many disadvantages so I wanted something better and I've got an idea recently which I would like to discuss here.

While trying to avoid the frame solution, we discussed the possibility of a "dead man's switch" in the form of an ajax call. After some tests we came to the conclusion that it wasn't possible due to the increased load, but it led me to another idea.

My current idea is to monitor the time between leaving one page of the application and requesting for another. If this time is bigger than some sane limit, it's obvious that user has visited something else and the application should kick him out.

The process and techniques are as follows:

  1. When the javascript's onbeforeunload event is fired on the client side, perform an ajax query which tells the web application that user is leaving a page. On the server side, store the time of this query into the session.
  2. Place a filter in front of the whole application. This filter compares the time of the current request with the last ajax request time stored in session. If the difference is bigger than some limit, invalidate the session for the rest of the application.

This is just a skeleton of the solution and won't work in many cases, but the implementation (at least in Java backend environment) is just a few lines of javascript and a very short servlet filter (which implements both - ajax time storing and time difference checking features), so it's worth at least identifying the pros and cons.

OnBeforeUnload support

The previous implementation requires the use of the browser with onbeforeunload support. It's supported in all major browsers (except for Opera AFAIK), but it's not a W3C standard and we don't want to rely on the non standard features, right?

Luckily a slight modification will help. The filter can be modified so that if there is no last ajax time at all, it assumes an unsupported client and lets the request go through. The security is decreased, but there is a choice now between security and accesibility (with enough security for majority of users).

Synchronicity of ajax requests

Important fact is the order in which ajax request and next page request comes. It's possible to manage order by using synchronous ajax, but it's generally not recomended technique. However assuming that request shuffling is very rare, a slight modification can help again.

By deleting a last ajax time after computing the transition time (so there will be always only one last ajax time at most), this system survives one pair of shuffled requests (if it's in accessible mode mentioned above). It's not very obvious, the picture would help, but I'm running out of time now, so you have to trust me:-).

Using of more browser windows with the same login makes this area more tricky, but more sophisticated logic should help.

What you can get

It's simple, this way you get your security-freak customer satisfied and maintain important aspects of web applications working, like back and forward buttons, bookmarks etc. You also don't have to pay much for that by the means of development resources or server load. So, what do you think?

Last update
201105220000