Xeroxing Personal Data From Your Browsing History

grease_boy writes "Xerox has filed a patent covering a technique to recover demographic information like your age, sex and perhaps even your income by analysing the pattern of web pages you browse. They want to license the technique to online advertisers and shops. Read the full patent here."

Comments

Hehe

[HomelessInLaJolla]

Because nobody could have ever thought of this before.

Re:More lame patents

[Brickwall]

Geez, anyone who read Peter Wright's "Spycatcher" - published 20 years ago - would realize that this patent is merely a specific implementation of the techniques Wright described in the book. In particular, he developed the idea of monitoring radio communications from the Russian embassy in London. All the transmissions were in code, so he and his colleagues had no idea of the content of the messages. However, Wright deduced that when the Russian "Resident" left the embassy, coded traffic between the embassy and his car would increase. By simply matching the comings and goings of various individuals from the embassy with increases and decreases in radio traffic volume, he and his team were able to pinpoint Russian intelligence agents promptly. (The whole book is a great read, by the way, and available at your local public library, I'm sure.)

In fact, I tried to get Nortel to implement something similar on their Meridian phone systems back in the early 90's. I thought that by tracking internal phone calls through "Call Detail Records" (CDR - which list the calling extension, called extension, date, time, and call duration), we could see patterns of calls between departments, and determine if repetitive patterns existed, such as many calls between sales and billing at specific times of the month, etc. I thought this might help identify operational issues, inefficiencies, etc. Of course, I was blown off.

Anyone want to give me a job?

Not too much to worry about

[PatentMagus]

Well, they aren't trying to patent statistical analysis or tracking cookies in general, but they are grabbing for a lot. The claims don't look too impressive either. All the claims are too broad without a single super narrow claim that might just get accepted. Maybe the person who wrote the claims didn't understand the math. Double click, some of the old shopping sites, and similar companies must have had this well before the 11/2/2001 priority date. There is nothing in the claims or the application saying it only applies across domains. The examiner is probably going to reject every claim and love it because it won't take too long.

1. A machine-implemented method for extrapolating user profile information from user web page access patterns, comprising:

computing bias values for a plurality of web pages; ancient and general

assigning said bias values to the plurality of web pages; ancient and general

detecting at least a subset of said web pages accessed by a user having an unknown user profile attribute; ancient and general - tracking cookies?

combining said bias values of said subset of web pages to obtain a combination result; and ancient and "combining" is way general

assigning a selected user profile attribute to said user in response to said combination result indicating a positive bias of the selected user profile attribute; ancient

wherein computing said bias values for the plurality of web pages further comprises determining a fraction of users with the selected user profile attribute who visit a selected web page as measured over the plurality of web pages.hmmm, intersting limitation. Calculating a demoographic from the tracked pages. Pretty ancient too.



Eventually, this app is going to land on a junior associate's desk who will try to rescue it. In the long run, there's a decent chance that Xerox will get a patent on a very specific algorithm applied to limited data that is gathered in a few different ways. Hey, inventions like electrostatic duplication don't come along every day.

It'll be used as smoke and mirrors

[Moraelin]

Aggreed, it is stupid, if your purpose is actually to do marketting with it and sell more products. But my guess is that it's not how it'll get used.

Thing is, if you think about it... it fits just neatly in the eternal 3-way total war, whee the ad provider tries to shaft both the advertising company and the web master, and in most cases the two try to shaft the ad provider too. Tons of useless metrics exist just so the ad provider can tell some company "here's why you owe us a big pile of money for serving your ads", or so they can tell the web master "here's why we owe you a pittance."

(And just so it doesn't sound like the ad provider is the only scumbag, the whole dot-com bubble was based on the "hey, look, we can rip off the ad providers" idea. Ad rates in the beginning were based on sites which had one banner on the whole site. It tended to be somewhat targetted too, since if the web master chose just one, it tended to be somewhat related to what the site was about. And people used to even click on them occasionally. And they were worth decent money. Then some people discovered, basically, "woo, but if we put 10 ads per page, now we're owed gazillions of dollars." Whole companies went to IPO with that as their only business plan. But I digress)

The fact is, the _only_ real criterion of whether a marketting campaign was successful, is whether you sold more stuff as a result. Everything else, eyeballs, clicks, etc, is just smoke and mirrors. It's just some useless metrics that get gamed all the time.

E.g., it may sound like "clicks" is a relevant number, but it is only in a world where everyone clicked only because you got them equally interested in the product. Once you figure out other ways to fake it, comparing numbers of clicks becomes apples to oranges. And once you give someone a criterion like "number of clicks" to justify their salary, we're already seen the result: fake UI ads, punch the monkey ads, and outright redirects served as ads. It doesn't mean that those people became more interested in a company's product just because they got hijacked, but on a "look how many people we got to click your ads" statistic it looks the same.

So my take is that this is what it will be used like. Some ad provider will make up some scientific-sounding "how well we matched your ad to a target demographic" metric and use it to justify why you should pay a premium to advertise through them. Never mind that the demographic was a wild guess, and it actually lost information in the process... twice. It will look neatly in the marketting materials anyway.