What It Takes to Truly Delete Data
fivethirtyeight.com
fivethirtyeight.com
It was difficult for me to learn this lesson personally. When I understood it, I deleted my Facebook account (best decision, btw), but I assume that even though my account was deleted there are still references to it all over. IDs in log lines, references not cleaned up -- enough to paint something of a picture of who I was and what I did with the site.
Now I'm more judicious in what I share, how I share it, which applications I give permissions. But I'm still leaking metadata everywhere. It's frustrating and terrifying that the cost of participating in our society is either a HUGE cognitive load in keeping an eye on all the data you emit or accepting that basically everyone will have full access to your (meta)data over time.
A few weeks ago, I walked away from a purchase because the store wanted to take my photo at POS. This was my first encounter with the store, I probably would not be a repeat customer, I just wanted to do a single purchase. Just take my damn money.
The stupid part is, I typically use the same credit card for every purchase. I know they could use that to query info product and know everything about me - and, I don't really care about that (I've given in to it). It's the constant asking for info by companies that I have no ongoing relationship with that I really hate.
Edit: Also, it makes me feel like some weirdo freak when I don't want to participate. It starts to cause a scene. Other customers are staring. My wife is embarrassed. Everyone is thinking, "just let them take the photo" or "the clerk is just doing their job".
The in store onboarding was actually a smart idea. Things like that are applauded by HN folks like us. Increasing LTV,... However, the clerk told me a pic was mandatory. I even said something like "take a picture of the wall then" and he said no like he was offended I asked him to cheat the system. I just turned around and walked out of the shop.
I was just trying to buy a tshirt. I didn't want to establish and ongoing relationship with this place photo or not.
I realize this request to reveal where you shopped, or what item you bought, is in response to your objection to being tracked, so I understand if you'd rather not say :)
Maybe, or maybe you made them second guess their own personal policy of what they are comfortable sharing. So long as you are confident I don't think a reasonable person would look down on you for protecting your privacy. I usually just say something like "I don't do that" when asked to share email/phone/zip with stores, never had a clerk get weird about it.
First thing that one needs to understand: you always leak data. It's a feature of reality itself. At the very least, every second of your life, the record of where you are and what you do radiates away from you at the speed of light. Of course interacting with matter, you leave stronger traces everywhere. Mechanical, thermal, accoustic. Most of the data isn't easily accessible in any meaningful way, but it doesn't change the fact it's always there.
Two: progress of technology increases our ability to access this data. There's no way to ensure that data is denied to all people other than stopping the development of new technologies. There's no meaningful way of preventing anyone from accessing the data that's available today other than rolling back the Industrial Revolution and going back to living in caves. Science and technology lets us see the physical world in more details, faster and cheaper, and it also lets us correlate information faster and cheaper. There's no way around that.
So that's about "metadata". As for "personally identifiable information", I may avoid giving some of it to random strangers, but I do assume that PII is public information. Because, again, there's no meaningful way to contain their spread. So pretending to yourself that any PII is private would be stupid.
> data isn't easily accessible in any meaningful way
and
> no meaningful way to contain their spread
If A is not readable then there is no need to contain it.
Now, I do carry a cellphone around so that exists. But, the vast majority of organisations have no access to that information.
When it does happen I've seen people post screenshots of the old tweet. This doesn't prove anything more than trivial knowledge of the Chrome Web Inspector.
I've thought about distributing a program so people can keep their own local archives of important tweets. There are plenty of ways to achieve this already, but that's besides my point.
How can we make it obvious when an important figure deletes a tweet without breaking Twitter's ToS?
That might be generous. They could have used Photoshop as well and had no technical knowledge at all.
1. The Twitter ToS prevents scraping the content in section 4.iii.
https://twitter.com/tos?lang=en
2. If you use the API instead, they have this rule:
> Only surface Twitter activity as it surfaced on Twitter. For example, your Service should execute the unlike and delete actions by removing all relevant Content, not by publicly displaying to other users that the Tweet is no longer liked or has been deleted.
https://dev.twitter.com/overview/terms/agreement-and-policy#...
https://projects.propublica.org/politwoops/user/realDonaldTr...
Back story:
https://techcrunch.com/2016/02/09/u-s-version-of-politwoops-...
Of course this does not apply to data that is actually being used/processed in cloud instead of just being passively stored.
Maybe a stupid question, but why don't they just write every bit to either 1 or 0 regardless of its current content? Are they trying to preserve some data and not others? Why do you need to do it multiple times if the bit being written is independent of the actual contents?
There is a concern that the write process is not perfect, so the old value of the bit might still be visible under a microscope.
Basically when a hard drive stores the data on the physical platter, it writes it as a series of concentric tracks. These tracks are spaced very close together. When data is written to the drive, the magnetic regions that are written on the track can "bleed over" to the intra-track regions, leaving a faint trace that with the proper equipment ($$$$) can be read, even if the tracks are all zeroed. Think of the magnetic regions as not being (say) "spikes", but rather "hills", where the greatest magnetic intensity is at the peak, tapering off as it reaches the edge/base of the hill. That energy might be small, but it is there.
By overwriting each track multiple times with varying data patterns, these leftover tale-tell magnetic leftovers are essentially pushed into the noise of the rest of the data, and are effectively "wiped". But you have to write the pattern multiple times, over all the tracks.
This effect has "recently" been used to an advantage (depending on how you look at it) for something called an "SMR Hard Drive" - or "Shingled Magnetic Recording":
https://en.wikipedia.org/wiki/Shingled_magnetic_recording
In this kind of hard drive, the tracks are purposefully placed very close together so that the magnetic regions purposefully overlap, allowing for a higher density on fewer platters (so, larger drives in a smaller space for a lower cost). However, they tend to be slower - and not recommended for purposes like RAID NAS or anything where you want speed and proper data retention (my understanding, at least).
What can bite you is the drive firmware silently remapping sectors, causing your write request to be mapped to a different sector altogether.
Edit: It looks like there is something called lorentz microscopy that uses an electron beam.
I'm also curious what's the third number they found in {0,1}