FridgeCam: A Late-Night Drinkies Story
October, 1998

Chateau 109D: Brad Chamberlain, Mike Perkowitz, Geoff Voelker, and Alec Wolman
Computer Science and Engineering, University of Washington

We have a refrigerator in our office that we stock with soda for the benefit of anyone who is thirsty. We've been doing this for about 5 years now without any problems - until recently.

We use a rather naive way to collect "donations" for soda: there is simply a box in the fridge that people drop their money in. When we run out of soda, we collect the money and buy more. Although the idea of an open box full of cash in a fridge that anyone in our department has access to might seem to be a bit risky, this honor system has worked without incident for years. Until the end of this past August, that is, when we started to suspect that bills were being taken from the box. Once our suspicions were raised, we started to watch the box much more carefully. By the beginning of October we were absolutely convinced that money was being stolen.

One evening while dining on Atlantic Street pizza we talked about who might be taking the money and what we could do about it. Since we had been doing this for so long, we couldn't imagine who had suddenly decided to use the box as their personal money tree. Since the fridge is beside Brad's desk, we were relatively sure that it was happening at night, but, beyond that, we had no clues. And even if we did have a likely suspect, we had no proof.

At this point, we could have accepted our losses (which weren't much, on the order of $40) and switched to using a locked change box to collect donations. Or we could have people keep running tabs so that they could give a donation every month or so. But we were fond of our current system, and we were very irked not knowing who was taking the money. So instead of changing the system, we decided to try and catch the thief.

We came up with a scheme that an office full of computer science graduate students would come up with: we pointed a camera at the fridge and took pictures all night long. We used the small camera on Mike's SGI Indy, and Brad wrote a script during the night that captured stills every few seconds [a sample still]. When dark, the file size of the compressed stills is small and easily filtered out. When there is a light source in the room (such as the opened refrigerator door), the stills are significantly larger, easily distinguished, and saved for review the next morning.

After about a week of taking pictures, getting the bugs out, and generally refining our technique, one morning we found a small series of pictures from the middle of the night. Flipping through the stills, it was clear someone turned on the light to the office and came to the camera and closed the shutter, all without being seen on camera [a three frame sequence of the office light on, the camera jostled, and the shutter closed]. Since the camera points at the fridge, it is straightforward for someone to sneak up on it and close it without being seen. Obviously our thief was more clever and aware than we first thought.

Our response was to throw more hardware at the problem. Brad devised a more clever scheme involving a second camera. Since we only had one Indy machine, we could only use one camera at a time, and so we used the second camera as a decoy. We got an extra camera identical to the one on Mike's machine, placed it where the original camera had been, pointed it at the fridge, and dropped the cable behind Mike's desk looking as if it were plugged in. We then hid the original camera on a filing cabinet under some books [like so], and pointed it at both the decoy camera and the fridge [a still from the hidden camera; the decoy camera is hidden behind the monitors on the left]. We also baited the fridge with three 1 dollar bills on top of a handful of change in the box.

After a few nights, we came in one morning and found the bills gone from the fridge and a sequence of pictures taken during the middle of the night. We were much more successful the second time around. We found stills of the light being turned on, someone going to the decoy camera and closing the shutter, then going to Brad's desk and looking under it, then going to the fridge, opening it, closing it while grasping something, and then walking away. We had our suspect, and some proof.

Who was it? See for yourself: [a still of the suspect closing the shutter on the decoy camera]. (Look for the tell-tale badge on the jacket...)

We then told the whole story to Ed, who eventually sent it to the campus police. That afternoon the captain and two detectives came by the office, we explained what had happened, showed them the fridge, the box, the cameras, and the pictures that we had taken.

The police set up a sting operation two nights later to try and catch the suspect in the act, but the suspect never showed up. They then confronted the suspect, who confessed to taking the money that one night, but not to any others. Currently, the campus police are pursuing charges.

(As an aside, when we were giving our statements at the police station, we noticed that the campus police detectives have binders on their shelves with much better sounding names than our binders. Instead of "SCSI II Specification", they have "Basics of Death Investigation".)

Perhaps, after reading this story, you might be wondering whether we still keep the money in the open box. You are welcome to take a look...just remember to smile.

Chateau 109D