Mar 10, 2008 04:30 AM
Jim Coyle
If her gig as Ontario’s Information and Privacy Commissioner doesn’t work out, Ann Cavoukian probably has a career as talent scout or matchmaker to fall back on.
Cavoukian reported recently on the proposed expansion of video surveillance on the TTC. A $21-million network of 11,000 cameras passes muster, she said, if there are adequate privacy safeguards.
“How do you make sure the system is controlled so you can prevent the kind of abuses that people are fearful of?”
Well, she found part of the answer in a research lab at the University of Toronto’s computer engineering department.
There, the work of doctoral candidate Karl Martin and professor Kostas Plataniotis got her so excited she plans to march off to Premier Dalton McGuinty to get them all together. “It’s fabulous,” she enthused. “It’s very cool.”
It can’t be every day that folks who produce masters’ theses with titles such as Embedded Wavelet-Based Coding of the Shape and Texture of Arbitrarily-Shaped Visual Objects win raves worthy of a new U2 release. So, Karl Martin will happily take it.
The 29-year-old researcher should stand as a calming example to teens (and their parents) worried about being unsure where learning will lead.
Now, he works in digital image/video processing and coding. But in high school, at Claude Watson School for the Arts in Toronto, he was a music major - albeit one with a “fundamental curiosity about how everything works.”
As a kid, he took apart “anything that I was allowed to” and remains “a closet Trekkie.” After high school, he applied to “the most challenging undergraduate engineering program I could find.” By third year, however, “I still wasn’t sure exactly what I wanted to do.”
Then, he was introduced to a mathematical tool called “wavelets,” did his thesis under Plataniotis, and the two saw the potential in security-related applications.
What they’ve come up with is a counterbalance to the inexorable expansion of camera surveillance into private and public spaces.
With their “object-based encryption,” a face or body on surveillance video is obscured. The background is visible. But only those with the decryption key, and a legitimate law-enforcement need, can access images.
“That’s the beauty of this,” Cavoukian said. It eliminates invasive use, but permits “legitimate law-enforcement use after the fact if it’s justified.”
It sounds as if Martin, math proficiency notwithstanding, has retained both an artsy side and sense of humour.
He’s a photographer who’s posted on his website some glorious shots of, among other things, sunbeams pouring ethereally through the windows of St. Peter’s Basilica. Those emailing him, moreover, are advised that “I am rather busy, so PLEASE be funny.”
Cavoukian was delighted his technology will be tried out at a TTC station.
“Once that happens, and we’ve got the footage and it’s working, then we’ll go to the premier and say, `Look, premier, you have this wonderful Ministry of Research and Innovation and this would be an ideal candidate for a technology to actually commercialize.”
Applause for Cavoukian’s report promptly arrived from two U.S. privacy experts, one of whom forwarded it to the Department of Homeland Security, recommending it as a blueprint for camera surveillance in public spaces.
For a young researcher, to have the influential promotion of Cavoukian and friends must be “very cool” indeed.
