Research

Managing Privacy During Ad Hoc Collaboration

What is it?

Marvin is at work browsing the web at his desk as he eats his lunch. He's unhappy with his job and registers at a couple of job search sites and peruses the on-line 'help wanted' section of his local newspaper. Later, his boss sits down beside him at the computer and wants to look at the competition's web sites with him again. His boss grabs the mouse and opens up the web browser's history files as they had been looking at these sites a couple of days ago. Marvin is uncomfortable and hopes that his boss goes directly to the sites from two days ago and doesn't notice his recent job search.
This research introduces the concept of privacy issues related to the incidental viewing of traces of previous activities during ad hoc co-located collaboration and/or sequential computer usage. Web browsers are used as the representative application in this research, as several of the convenience features record traces of previous web page visits.

Survey of Web Browser Privacy Issues

We are currently analyzing an on-line survey with 155 users to explore the dimensions of privacy related to the incidental viewing of previous activity that arise during face-to-face collaboration and in circumstances when computers are communally used. Participants were recruited from business, the general university community, and the public. Participants were grouped according to their primary setting of web browsing activity (home, school, work) and the type of computer used (laptop, single user computer, shared computer). We are investigating if there are different levels of privacy desired depending on the user and how the context (their relationship to the viewer and the level of control the viewer has) of subsequent viewing of private information impacts privacy concerns. We are also examining the influence of device mobility on privacy concerns and the actual settings in use for the various browser features.

Privacy Gradients

In order to enable classification of visited websites, we require a common terminology. We introduce a four-tier privacy gradient scheme that partitions web sites: Public, Semi-Public, Private, and Don't Save (see Figure below). If a site is something that you would like to access again, you would want traces of it to appear in your browser convenience features. These traces should be stored with some associated privacy level. Public sites are those that you are comfortable with anybody and everybody viewing, including the Queen of England (hence the crown in the Figure). Private sites are those that you would be comfortable with only yourself and possibly a couple of close confidants or a spouse viewing, people with whom you share just about everything. Semi-public sites fall somewhere in between: depending on the context of the viewing, the pages would be considered to be public or private.

Field Study 1

We introduced the above 4-tier privacy gradient to allow 20 study participants to classify privacy levels associated with their actual web browsing over the course of a week-long diary study. A browser helper object was installed on their laptop computer so that the majority of their web browsing could be captured. Participants used an electronic diary (see below) to classify each visited page using the privacy gradients. Data captured included the date/time stamp, browser window ID, and privacy level for each web page visited.

Results include analysis of the privacy comfort levels of individuals, their current privacy management strategies, their browsing behaviours, and their use of the privacy gradients. This initial exploratory study provided important insight that will guide the development of a privacy management system. Results were reported in short talk (privacy gradients) and a poster (web browsing today) at CHI 2005.

Field Study 2

We conducted a second field study to gain more information about the role of context of web browsing behaviour (location of browsing, type of computer, type of pages visited) across a broader demographic. The methdodology was similar to that of the first field study, but in additional data captured included the page title and URL, location of use, web browser focus events, and number of unique pages visited. Participants were able to blind the page title and URL if desired. Analysis is currently underway.

Publications