Perimeter Security and the Great D.C. Porn-Surfing Scandal

The story of the dismissal of 9 District of Columbia workers for surfing pornography on the Web from their work computers is both a social story and a technical one.

My first thought, when I see a story like this, is how stupid some people can be. But it's even worse than it looks.

"Each of the nine employees clicked on porn sites more than 19,000 times last year," according to the city's internal investigation. If you divide this up and assume a 225-day year (probably a lot for city workers, who we can also assume don't come in much on their days off), it comes out to just over 84 a day.

But 19,000 is just the floor for this group; the Gold, Silver and Bronze go to three workers with over 39,000 "clicks," for an average of 173 per day. What about the employee who only clicked 18,000 times in the year? Those between 2,000 and 19,000 will receive letters of reprimand or suspensions. Under 2,000 (not quite 9 clicks a day) and it's not clear what, if any, consequences there are.

But wait -- there's more.

The District actually had software in place (from WebSense) to monitor such behavior. It appears, from the Post story, to have been client-based software, and was only installed on 10,000 of the District's approximately 30,000 computers. Was the investigation the result of alerts from that software? Maybe, maybe not.

"The investigation was launched late last month after D.C. Chief Technology Officer Vivek Kundra received a tip from an employee in the Office of Property Management," officials said. So it sounds like no standard procedures were in place to act based on the results of this monitoring; it took a tip to the CTO to do it. And since only one-third of the computers were investigated, it's not unreasonable to assume that only one-third of the offenders were caught.

This story is, to me, a great example of the necessity and advantages of perimeter-based security. Imagine trying to manage an application on 30,000 desktops! Of course, desktop protection (these days the term has become "endpoint" protection) is necessary for many purposes, but it's a perilous approach doomed to fail if connections are not monitored at an earlier stage.