Google gives away face recognition technology for free. But what happens if this technology becomes part of regular web search? Is there any privacy left?

The image oriented part of the social web is dominated by Yahoo’s Flickr.

Google also has its own image tools, however: There is the Picasa software for Windows and Linux, a free photo organizer, and Picasa web albums, an online photo sharing service like Flickr.

NRKBeta, a blog run by The Norwegian Broadcasting Corporation, explains how the newest version of the Picasa Windows software opens up new possibilities: It is, to a reasonable extent, able to recognize faces.

Google puts it this way:

Instead of tagging your photos individually, you can quickly identify and label many photos with one click.

• Name tags help you automatically find similar faces in your photo collection. All you have to do is enter a name or choose from your contacts.
• Once you’ve named the people in your photos, you can do things like sort your photo collection by person, create custom slideshows, and easily share photos with the people in your albums.

Picasa Web can automatically go through your pictures and identify the pictures it believes to be of faces. Then it will try to sort them into different piles of pictures, one for each person. Then you tag these pictures with the correct name.

Picasa is not always successful in identifying the right person, but all in all it does a good job. And given the amount of digital pictures people keep these days, some help in tagging them is most welcome.

Google is using the technology of Neven Vision, a company it bought in 2006.

Making face recognition part of regular image search

The French search engine Exalead lets you restrict image search results to faces, as does Live.

None of them lets you search for a particular person, however. Nor does Google Image Search.

Yet.

But the technology is there. It is a bit rough at the edges, but it works. In principle it should be possible to develop an image search engine that lets you upload a picture of a person and then helps you find pictures of that person on the web.

It would require a lot of processing power, but processing power is becoming less and less of a problem every day.

A huge privacy problem

Today you may use Google Image Search to search for a person.

Google and the other image search engines make use of the image ALT tag, the link text and the surrounding text on a web page to determine what the picture is of.

If there is no text, Google cannot find it.

With face recognition however, the number of hits will increase tremendously and thus the possibility finding an incriminating photo.

Given that more an more people upload their photo albums to the Net, the number of pictures that might — on purpose or accidentally — include you will increase as well. Moreover, more and more cameras will have their own GPS and a clock, tagging the pictures automatically with place and time.

When that happens you may ask questions like these:

• Where was NN on Monday morning? Search for images of him/her tagged with that date.
• Who took part in the protest meeting outside the Iranian embassy? (We are sure Iranian intelligence would love to find out.)
• What was dad doing outside a seedy nightclub Saturday night? (Ouch…)

You could even, given the right tools, track a person’s wanderings from one part of town to another — at least for areas well covered with tourists snap shots. We take it for granted that surveillance camera footage will not be included, but live web cameras might.

All right, so most of us will not bother with this. But foreign intelligence might, private detectives, surely, and our own law enforcement, definitively. But do we in the general public want this?

We guess the search engine’s lawyers are already struggling with the problem. For Google an web wide version of face recognition search might lead to a PR backlash, public protest and trigger happy politicians aiming for the “new Microsoft”.

We are, indeed, living in interesting times.

See also: Are there search engines for image, sound or movie content? by Lars Iselid.
Revamped Google Picasa site identifies photo faces (Underexposed Blog September 2 2008)