4 ms·

So is the claim here that even though they dont know me at a personal level, they know me as ”shopper_1092716718” and can target me whenever they want? If so,

So is the claim here that even though they dont know me at a personal level, they know me as ”shopper_1092716718” and can target me whenever they want? If so, another fantastic example of expensive and complex regulation being gamed by the big players while the smaller entities are left with the cost.

INTPenis 6y ago
In Oslo a while back there was a faulty billboard that started outputting all its log data to its main screen. It was essentially tracking everyone who passed it by characteristics like gender, if they had glasses, where they were looking. But yeah, every person was just an entry. There was no way for it to know who that person was. I wish I could find the story but my google-fu isn't helping.
- whatok 6y ago
  https://theoutline.com/post/1528/this-pizza-billboard-used-facial-recognition-tech-to-show-women-ads-for-salad https://theoutline.com/post/1528/this-pizza-billboard-used-f...
- tyfon 6y ago
  That's where people are wrong, it's not about not identifying the 99% that share similar traits, it's about not identifying those who do not share traits with others (in that area). Such a billboard in a small town up north in Norway could easily identify the individuals based on those characteristics since the population is so low.
Tarq0n 6y ago
This could be fine as long as there are very strong regulations against de-anonimizing the data (including jailtime for violators) and there was a regular rotation of the keys used for anonimization. People's privacy usually isn't impacted until datasets are combined, compromising pseudonimity.
- raxxorrax 6y ago
  I think you should basically kick advertisers in the nuts for stuff like this. I doubt it would be legal in my country and if it was, I wouldn't mind doing something illegal to the billboard. Can always use new cameras anyway.
- waihtis 6y ago
  Does the de-anonymization serve any purpose at this stage? If the advertiser can successfully profile me based on what they can gather completely anon, I don't see why they need the further info. However, for any smaller player wishing to compete on the advertising side the cost of entry is now humongous because of intense up-front cost to setup capability that is in compliance. Et voila, corporations win in the end.
  - ourlordcaffeine 6y ago
    I'm guessing de-anonymised is more valuable, as it can be linked with other datasets to create a more in-depth profile, then sold to parties interested in knowing everything about a particular person.
    - tremon 6y ago
      Yes, but that definition is circular: if the dataset is detailed enough that a particular person can be linked across multiple datasets (with X degree of confidence), it isn't anonymised. So what you're really saying is that a detailed dataset is more valuable, and shouldn't be considered anonymized. Or, more succinctly, "anonymized" should not be boolean, but is a quantitative property of the dataset.
      - Tarq0n 6y ago
        Anonymization is a bit of a misnomer here. This dataset is clearly pseudonymised. Most datasets at the level of the unit of observation are never truly anonymous, unless they barely contain any real information at all. However this can be sufficient to protect privacy as long as the dataset is never allowed to be combined with other information.
tomp 6y ago
I think we need to change the definition of "personal data" in GDPR from things like name, phone number etc. to just bits. It takes around 33 bits to identify a person, with some leeway (so maybe like 26 bits). Sex == 1 bit. Hair color =~ 2 bits. Glasses/no glasses =~ 0.6 bits. Phone number == 33 bits. Location to 1km precision would be many more bits in a rural area (less people / lower density) than in a big city. And so on. This would take care of the metadata. It would also put the companies on the defensive - if _the company_ thinks they're collecting just 20bits of data, but if I can do some fancy ML on their data (require regular external audits of collected data if more than say 10k people were tracked) and retrieve 27bits, that would put them in breach (& massive penalties).
- hyperman1 6y ago
  They more or less already do: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32016R0679 https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CEL... Chapter 1 Article 4, definition of personal data [...] an identifiable natural person is one who can be identified, directly or indirectly [...] or to one or more factors specific to the physical, physiological, genetic, mental, economic, cultural or social identity of that natural person; If I read this correct, if you can be identified by a tuple (sex,hair color, classes, more attributes), the GDPR applies, even if they have no name etc... Hence, even an anonymized ID is still identification for the GDPR.

INTPenis 6y ago

In Oslo a while back there was a faulty billboard that started outputting all its log data to its main screen. It was essentially tracking everyone who passed it by characteristics like gender, if they had glasses, where they were looking. But yeah, every person was just an entry. There was no way for it to know who that person was. I wish I could find the story but my google-fu isn't helping.

whatok 6y ago
https://theoutline.com/post/1528/this-pizza-billboard-used-facial-recognition-tech-to-show-women-ads-for-salad https://theoutline.com/post/1528/this-pizza-billboard-used-f...
tyfon 6y ago
That's where people are wrong, it's not about not identifying the 99% that share similar traits, it's about not identifying those who do not share traits with others (in that area). Such a billboard in a small town up north in Norway could easily identify the individuals based on those characteristics since the population is so low.

Tarq0n 6y ago

This could be fine as long as there are very strong regulations against de-anonimizing the data (including jailtime for violators) and there was a regular rotation of the keys used for anonimization. People's privacy usually isn't impacted until datasets are combined, compromising pseudonimity.

raxxorrax 6y ago
I think you should basically kick advertisers in the nuts for stuff like this. I doubt it would be legal in my country and if it was, I wouldn't mind doing something illegal to the billboard. Can always use new cameras anyway.
waihtis 6y ago
Does the de-anonymization serve any purpose at this stage? If the advertiser can successfully profile me based on what they can gather completely anon, I don't see why they need the further info. However, for any smaller player wishing to compete on the advertising side the cost of entry is now humongous because of intense up-front cost to setup capability that is in compliance. Et voila, corporations win in the end.
- ourlordcaffeine 6y ago
  I'm guessing de-anonymised is more valuable, as it can be linked with other datasets to create a more in-depth profile, then sold to parties interested in knowing everything about a particular person.
  - tremon 6y ago
    Yes, but that definition is circular: if the dataset is detailed enough that a particular person can be linked across multiple datasets (with X degree of confidence), it isn't anonymised. So what you're really saying is that a detailed dataset is more valuable, and shouldn't be considered anonymized. Or, more succinctly, "anonymized" should not be boolean, but is a quantitative property of the dataset.
    - Tarq0n 6y ago
      Anonymization is a bit of a misnomer here. This dataset is clearly pseudonymised. Most datasets at the level of the unit of observation are never truly anonymous, unless they barely contain any real information at all. However this can be sufficient to protect privacy as long as the dataset is never allowed to be combined with other information.

tomp 6y ago

I think we need to change the definition of "personal data" in GDPR from things like name, phone number etc. to just bits. It takes around 33 bits to identify a person, with some leeway (so maybe like 26 bits). Sex == 1 bit. Hair color =~ 2 bits. Glasses/no glasses =~ 0.6 bits. Phone number == 33 bits. Location to 1km precision would be many more bits in a rural area (less people / lower density) than in a big city. And so on. This would take care of the metadata. It would also put the companies on the defensive - if _the company_ thinks they're collecting just 20bits of data, but if I can do some fancy ML on their data (require regular external audits of collected data if more than say 10k people were tracked) and retrieve 27bits, that would put them in breach (& massive penalties).

hyperman1 6y ago
They more or less already do: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32016R0679 https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CEL... Chapter 1 Article 4, definition of personal data [...] an identifiable natural person is one who can be identified, directly or indirectly [...] or to one or more factors specific to the physical, physiological, genetic, mental, economic, cultural or social identity of that natural person; If I read this correct, if you can be identified by a tuple (sex,hair color, classes, more attributes), the GDPR applies, even if they have no name etc... Hence, even an anonymized ID is still identification for the GDPR.