Australian Bank Anonymizes and Releases Billions of Records

I punched in my details and there are approximately 100 people like me. What has occurred to me is that people like me (20's, single) that live in rich mobile areas of Sydney spend less than the national average across most things, and overall. But also, compared to people like me, I get away with about 1/8th of what everybody else spends, and do not feel poor. To the contrary in fact.<p>What is going on here? I am inclined to believe the data.<p>What this has really reinforced for me is the extent to which other people must spend their income as it comes in. It has also reinforced the importance of my friends, connections, values &#38; personal knowledge. My friends who enable me to live a life that feels rich whilst, apparently, spending about 1/8th of what everybody else does.<p>I really am surprised. I was not raised frugally and if I want something and have the money, I generally buy it &#38; I don't need more than I have.

Given the history of people being identified from "anonymous" records (thanks a <i>lot</i>, information theory) ... well, I'm glad I'm not banking with the NAB.

For what its worth they have not "released" the data. I cannot copy any of the text from the FAQ section but if you browse the FAQs and "about this site" it looks like the whole site is basically an ad/demo of Market Blueprint[1]. But I could be wrong. estromiund, do you know where they released the data?<p>[1] <a href="http://www.quantium.com.au/market-blueprint" rel="nofollow">http://www.quantium.com.au/market-blueprint</a>

I wonder how accurate this actually is... I punched in my details and apparently for my details there are less then 10 people like me. Considering where I live, age etc... that seems unlikely.

If they didn't acknowledge the well-known risks and how they got around them, they will have some concerned customers and questions from regulators.<p><a href="http://en.wikipedia.org/wiki/AOL_search_data_leak" rel="nofollow">http://en.wikipedia.org/wiki/AOL_search_data_leak</a><p>Robust De-anonymization of Large Sparse Datasets <a href="http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf" rel="nofollow">http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf</a> Regarding Netflix's competition: "We demonstrate that an adversary who knows only a little bit about an individual subscriber can easily identify this subscriber’s record in the dataset."

Interesting, as a customer of ubank I was concerned with the concept, especially after the data deanonymization techniques developed post AOL and Netflix.<p>However after using the site and entering my demographics (all public properties) I could see that I was like 56 people in area; their base spending patterns did not reflect me at all and I felt like a snowflake. Sometimes big data makes you feel special.<p>At no time was I shown transactions, merely aggrigate figures in categories. No privacy issue here, keep being decent and ethical national bank!

I'd say the initial 6 questions are modelled on Census data, not anonymous bank records.

Is anyone aware of what method they used to make the data anonymous? I'm currently doing research in data masking tools and would be interested to know about the techniques used and the performance of the tool.

Kind of cool, but I bank with the bank in question (Ubank / National Australia Bank) and I'm not sure I'm completely comfortable that they'd use my transaction history like this.

I put in my numbers to see what it would be like living on the other side of the world (I'm in Dublin, Ireland - it is currently 9C and raining) using the Sydney postcode 2000.<p>It says my expected house &#38; home costs are $2000, which seems rather high for me. In London and here in Dublin, pretty much everyone my age (mid twenties) is in a houseshare, so we would pay that for the whole property, but split it three ways or so. Is housing just that much more expensive, or what?

If we go by the history of such anonymization attempts (AOL, Netflix) it is only a matter of time that someone figures out how to get PII from this data.

Spooky. Tried it out and found it's prediction accurate for brands and companies I might use. Which was disappointing in a way, I was hoping I might find something new to try. Instead I'm well, just like me....

Nice idea but an absolute terrible website. E.g. If you go into the about section and expand one the questions you get this weirdo scroll bar. Seems like they're forcing the layout into a fixed dimension.

You can see where rich people eat.