Concerns Over Microsoft's Internet User Profiling

jcatcw writes "Microsoft research on Internet user profiling could lead to tools that help repressive regimes identify anonymous dissidents, the Reporters Without Borders advocacy group warned last Friday. Microsoft's new algorithms correctly guessed the gender of a Web surfer 80% of the time, and his or her age 60% of the time. "In China, it is conceivable that this type of technology would be used to spot Internet users who regularly access such 'subversive' content as news and information websites critical of the regime," the group said."

Comments

If I were head of an oppresive regime...

[Anonymous Coward]

I'd be uneasy about partnering with a bunch of totalitarian control freaks like Microsoft.

Re:If I were head of an oppresive regime...

[Anonymous Coward]

That said, I reckon I could guess the gender of a web surfer 80% of the time.

if (site.equals("slashdot.org")) male = true;
else if (site.equals("cutepuppydogs.com")) male = false;

oh noes!

[Anonymous Coward]

Unless you happen to be the only 20-25 year old male in China I think you're safe.

Re:oh noes!

[Short Circuit]

Microsoft's new algorithms correctly guessed the gender of a Web surfer 80% of the time, and his or her age 60% of the time. "In China, it is conceivable that this type of technology would be used to spot Internet users who regularly access such 'subversive' content as news and information websites critical of the regime," With that kind if inaccuracy, it could hardly be considered a reliable indicator of identity. It seems much more tuned to generating demographic data.

Not that China couldn't use that in targeted propaganda...

Poor accuracy

[mws1066]

80% and 60% are both actually very poor accuracies. I wouldn't be worried; this won't be taken seriously as any type of reliable profiling.

Re:Poor accuracy

[Lockejaw]

I am worried about such low rates. People seem inclined to believe anything a sufficiently large computer says.

Re:Poor accuracy

[ajanp]

That same day, Erik Bratt, a Microsoft marketing communications manager, fired a preemptive salvo about the company's age- and gender-guessing research. Bratt first downplayed the research results, saying: "[The researchers] actually found that they could not, with a high degree of accuracy, predict age from Web browsing activity." He also swore the Redmond, Wash. developer off using the resulting algorithms. "Microsoft currently has no plans to use the capabilities found through this research in our products and services," Bratt said. So... they tell people that they have the ability to kind of, maybe, sort of, predict age and gender, but that their current algorithms are basically just a bunch of BS. Then they decide to also mention that despite the fact that the foundation for this technology doesn't really work accurately and the have no plans to use it in any products, they're going to continue working on it anyways.

Microsoft's researchers said they would expand their work to other demographic attributes, such as occupation, educational degree and geographic location I mean, ofcourse that's the logical thing to do so that they could hopefully get it to work more accurately, but why mention that you're even working on the technology if you're going to immediately dismiss it by saying that it doesn't really work.

Then again, now I guess it solidifies the opinion about why Microsoft is really so sore about the Google-Doubleclick deal http://slashdot.org/article.pl?sid=07/04/16/021720 3. Google's got a leg up on what they have apparently already been working on and now they're at a disadvantage.

Re:Poor accuracy

[reddburn]

80% and 60% are both actually very poor accuracies. I wouldn't be worried; this won't be taken seriously as any type of reliable profiling. It depends on the context. If you were talking baseball, it'd be a pretty damn good batting average. It's also good enough to get a warrant, and good enough for a grand jury to indict, and those latter two are what should worry you.

Re:Poor accuracy

[MyOtherUIDis3digits]

I'm already doing my part. While my surfing habits are definitely male-oriented, I'm sure they would determine me to be 14-16 tops.

If you were a totalitatian regime

[supersnail]

... wouldn't it be easier to look up the IP address and persuade the ISP to hand over the user details?

The old ways are often the best.

Here's my algorithm:

[erroneous]

If you read this post you are probably male. (80% correct)
You are probably in the 20-35 age group. (60% correct)

(I know I'll only get negative responses to this post of the type "I'm reading this and I'm a 47-year-old woman!" That's Ok. You're in the other bracket.)

My algorithm is as good as Microsoft's. Can I have a research grant please?

Re:Here's my algorithm:

[spellraiser]

You got me pegged. Curse you!

Can I have my privacy violation lawsuit settlement, please?

Not too bad for now...

[taupin]

Printable Version

Right now this doesn't worry me too much - after all, how much "identification of anonymous dissidents" could someone do based only on one's gender and a rough estimate of age? On the other hand, if Microsoft do expand to geographical location, occupation, and educational degree as mentioned, then it's rather worrying.

O Rly?

[vivaoporto]

First things first: why China? (The same question applies to Venezuela, Russia, Brazil or whatever is the target of the Slashdot "fifteen minutes of hate" of the day). Of course people should be concerned about what these countries do wrt losses of privacy and basic rights, but what about U.S. and E.U.? As we talk, they are working on a new agreement to share data from passengers on trans-Atlantic flights, a much more effective way to profile people, because it contains name, address, gender, destination, credit card number, everything, without needing to make any kind of assumption, everything is plain and clear. This is why I think that not only "in China", as the summary states, but in most countries in the world, this information can and will be used to tag people indiscriminetaly, subversive or not, terrorist or not, law abiding or not. So, take care of your own backyard before to point the poison ivy in your neighbor one.

Second, it is not like if Microsoft was the only one researching and developing on this field and, more than that, it is not like if Microsoft was not researching on this field, any government interested on this kind of technology would not research itself, or fund research on its public universities. So, throwing Microsoft name on the mix only reinforces my point, this submission is nothing but a flamebait, being the flame targets the usual suspects, proprietary software and communism.

Re:O Rly?

[bahstid]

think of the child^H^H nese!

The Usual Suspects

[blueZhift]

This reminds me of the scenes in Casablanca where the police are told to round up the usual suspects. Ultimately the accuracy doesn't matter to the government anyway. Worldwide, I think we are moving in a direction of less freedom rather than more, spearheaded by wrong-headed anti-terrorism hysteria in the US. So why should they care about accuracy, they'll just round up whoever fits the profile and sort it out later, or not.

I know who it is!

[Phat_Tony]

Right, so they're about 50% sure that it's someone who's both male and aged 24-30, living in China. It should be easy to pick out the individual from there.

Re:Extension time methinks

[alexhs]

What about somebody writing a browser extension that performs bogus searches in the background, for no better reason than to frustrate "profiling" attempts? Is this feasible? Already done