Slashdot Mirror
Google to Anonymize Users' Search Data
Google's official blog states they are on an effort to anonymize their search data after 18-24 months. After previously fighting turning over search data to the feds, it looks like they are striking another blow to the "think of the children" crowd. Any bets on whether MSN or Yahoo! will follow suit?
Why would Google have to comply with EU regulations? :?
Maybe because they do business in Europe?
My site
My guess is they don't do it immediately is because there is internal business value in mining the data. User patterns, length of stay, etc. After 18 or 24 months, the internal value has dropped significantly as things change quickly. I would have thought that the value would have dropped even quicker then that, say after 6 months or maybe a year.
I never got why google needs to keep all that history without anonymizing it.
There is - as far as I can see - no rational argument that has to do with improving search results because you have them tied to individuals.
And yes, keeping tabs on half the globe is evil too...
MP3 Search Engine
And anonymous proxies do not need to make money or provide much of a service unlike google, logs are very useful for such things.
Google plan to make it "more anonymous". Like pregnancy, data either ARE anonymous or they ain't. You can't qualify an absolute, and "anonymous" is an absolute condition indicating lack of information.
All they have to do is erase the logs every day or just not keep them. It doesn't "take an effort". Anonymous proxies have been doing this for years.
I know where you're coming from, but that would kinda fuck with their targetting advertising business model dontcha think?
There are shills on slashdot. Apparently, I'm one of them.
If you've got nothing to hide, you should have no problem with this.
I'm between the two extremes of agreeing with you and agreeing that data needs to be retained. As any of us who have taken a statistics class (or four) can tell you, you don't need access to the whole sample to provide accurate data. So, say, for instance, the Google engineers were working on a specific niche of the web, say, dog lovers. If I were designing something to better suit dog lovers, my first step would be pulling a report on the common search patterns of people that search for dog-related topics.
Historical data that identifies a unique user is extremely useful. I do the same thing with our Intranet search and report tools. If I want to improve something, oftentimes the logs will give a very telling tale. (This accounting department employee searched for "expense", then "expense excel", then "expense spreadsheet", then "expense log", finally getting his document. I can then add the keywords 'excel' 'spreadsheet' to the actual document entry.) That said, you don't actually need to know who the unique user is, for all intents and research purposes, User5486734067 is just as useful as an IP+Cookie.
This is why it pays to have a modicum of computer knowledge.
Assuming you're not trolling...
When you send a query to google, it goes over the "internet" in the clear. That is, not encrypted. Anyone who can see it can read it. Well who can read it? Turns out a lot of people. Between me and google are probably 10 different boxes. 5 of which are just my ISPs routers. The other five are boxes on other networks, not even related to Google.
There is no inherant requirement for privacy like there is with telephones (maybe their ought to be one). But that said, you're giving your data to Google, willingly no less. That gives them every right to record it. You gave them permission by using their service, I guess you never read their TOS which is your fault, not theirs. Think about the analogy in the real world. This is like you handing your drivers license to every stranger you meet, then getting upset when some of them write it down.
If you don't want your assets [IP, location, name, platform, etc] leaked to Google you should use an anonymous proxy.
Tom
Someday, I'll have a real sig.
List of nifty little phrases that have bitten their speakers in the ass:
Now Google brings us:
Let's just be less evil, now that we've been caught.
Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong fix.
There's 2^32 IP addresses under IPv4. If Google is doing the hashing, then they know the hash function. How long do you think it would take them to brute-force break the hash by hashing every possible IP address and creating a map from the hashed values back to the originals? Express your answer in microseconds.
(If your solution is to increase the space of inputs by adding a variable salt value, please explain how this allows them to use the resulting hashes for aggregation.)