I don't see how this follows. Your initial implementation of your idea was not very effective. You learned that you needed faster hardware. You implicitly discovered that you wanted something different than you initially thought you wanted. (In any kind of construction, including software architecture, the change orders that result from learning what is wrong with the design while building the design are both expensive and important.)

I suggested one way of solving the problem: with slightly more expensive hardware and a shift to already-available software that solves your problem, which I trust (1) because it is all free software so everybody can see what it does by reading it, which you can't do with the code running on the other side of someone's service, as you remark; and (2) because the project making it is run by me and my comrades, and we trust one another to follow the rules of transparency and respecting users' rights, subject always to complete ongoing inspection.

But we could also solve your problem at no hardware cost and using only free software. We could use the storage that Columbia gives your for free on your cunix account, and we could use tools like "tomb" to create an encrypted container on that account and tools like sshfs to access that container over a secure connection from your laptop anywhere. The effect would be to give you a folder contianing all your files organized however you want them that would look like Google Drive does, but which would be secure at rest and secure in flight and operated by you without any hardware or any cost. The space is limited by what Columbia gives you, but we could do the same thing using an Amazon S3 bucket---which you wouldn't have to trust Amazon about because it is encrypted with keys they don't have---and s3fs or the equivalent.

In other words, it's good to have some amount of knowledge, and fairly illogical to say that you wanted to find out whether something was possible to learn, but because you weren't successful in learning it the first time you tried learning is impossible and everyone should know that your experiment proves they should give up.

Why does the answer to this question, which I taught and which we discussed, not make an appearance here? When one uses free software one does not have to "trust" people who tell you what software does: one can read it. And if you don't or can't read it, you can listen to the conversation among the millions of people who do make, use, improve, and distribute free software, from companies like Oracle and Red Hat in the S&P 500 to the technical workers in your social circle. They teach people how to understand how it works, and are constantly checking on its reliability. What you are posing as a recursive difficulty in knowing whom to trust is actually a full technical ecology of trust management that answers your objection so basically and so strongly that most of the world's corporate IT already depends on it.

Privacy and Trust

I was naïve to trust you, Google.

You know nothing, John Doe.

Can I trust you?

Why do you say these records are "anonymized"? That's the one thing they are pretty sure not to be.

Users who care about both convenience and privacy may be motivated to argue that they should have a say over data about themselves. Is this position well-justified?

What does "well-justified" mean? Legally, ethically, logically?

Generally speaking, you would want to begin an essay with an idea, not an open-ended question. The reader needs to know why she should read the essay. Obviously, if the reader is a philosopher who doesn't care about time, whether or not to have an idea may be precisely the stage at which she likes to begin reading. But for the remaining fraction of humanity, you haven't provided a reason to read the essay, and that means it will most likely not be read.

Once again the unclarity about what sort of argument is being rehearsed makes reading difficult. One sentence seems to be discussing a legal concept, ownership, while the next appears to be stating a political principle, though whether the principle is that things other people don't know we should have some say in controlling, or whether we have some stake in controlling things others know too is left vague.

Cannot, why? What is "perfectly anonymized" data? Why is data always either about one person or about nobody? If someone is keeping track of all Jews, all Muslims, or all Tutsi, that's not about privacy because privacy is only about one person at a time? Perhaps the definition of "privacy" is at fault. But we don't know what that definition is because you haven't given one.

I don't understand the idea of diluted responsibility. Is that like when three polluters each put one poison into a river?

This anecdote doesn't seem to me related in any way to the previous analysis. I do think that it would probably have been a good idea to consult somebody else rather than trying to do your engineering by the light of nature. By spending $150 instead of $90, you could have had a single-board computer much faster than a RasberryPi? , and if you had installed FreedomBox? software your personal cloud would have done a great deal more for you than it appears you could figure out yourself how to do, which isn't surprising considering that dozens of experts have already been working on it for seven years and you took a couple of hours.

But how did this help us to determine whether something or other was "justified"? And what was the idea the essay wanted to communicate to the reader? I think it was "privacy, whatever that is, isn't very important even though we intuit at first that it might be. But mostly it's just a lot of unnecessary trouble." Did I get that right?

Review of the Argument

A Way Out?

I cannot think of any files that I need remote access from time to time, that I am willing to tolerate the quirkiness of my personal cloud so as to prevent any service providers to harvest any data on me, and that are not sensitive enough for me to worry about targeted hacking. In the end, I do not know what to do with my personal cloud. I pulled the plug by the end of April.

Do users have a right to control data about themselves?

-- By LeonHuang - 05 Mar 2017

Service providers typically keep anonymized records of how users are using their services, and the users’ agreements, which the users have consented to, typically allow service providers to use these records. Users who care about both convenience and privacy may be motivated to argue that they should have a say over data about themselves. Is this position well-justified?

An Review of the Argument

The position appeals to our intuition in two ways. First, the content of the records is produced by the users. Users may deem themselves as the author of such data and in turn the owner of the data. Second, users expect to have control over their privacy. The records on users’ behaviors contain private information that users do not expect others to know, and therefore the users would think that they should have some say in how the records are being used.

There are counterarguments too. The keeping of the records is performed by the service providers. Although the content of the records is largely dependent on users’ behaviors, the service providers are nevertheless the author of the records in the same way that the people writing biographies are the authors of those biographies rather than the subjects of those biographies. In addition, anonymization alleviates the privacy concerns. After identity-sensitive information is striped from the dataset, the record of each individual’s behavior becomes less likely to violate that individual’s privacy. The data is no longer about each of the individual users but rather about a set of users sharing some demographic characteristics. Users cannot be justified to argue for more when the data about them is perfectly anonymized.

But perfect anonymization is a high standard. Stripping away identifiable information is not always enough. For example, the TLC Trip Record Data provides publicly available information on the dates, times and locations of all taxi pick-ups/drop-offs in the New York City in a given year. Although the data does not include the identity of the passengers, it nevertheless increases the risks of privacy violations when it is used in conjunction with other publicly available information. Celebrity cab rides become easier to identify. People may simply find a photo of a celebrity getting into a cab and use the date, time, and location to find out where the celebrity was going. And in turn people may easily find out where the celebrity lives. The average Joe faces the same risks. An acquaintance may easily find out where you live after seeing you leaving in a cab after work. This tension between privacy and highly-accurate geolocation data has led to a proposal that the TLC should reveal only census tracts instead of the exact coordinates.¹

Anonymized data is still capable of revealing key information about users’ identity when such data is used in combination of data from other sources. A link can be established between datasets when there are significant overlaps. Significant overlaps are not as difficult to achieve as one would expect. In the case of the TLC Trip Record Data, date, time and location provide the enough overlaps. After the link is established, the profile would be complete as long as one of the datasets contains the user’s identity.

A real privacy concern arises when multiple service providers give their purportedly anonymized records to one datacenter, where the complete profiles are being constructed. Datacenters like this already exist, although the profiles being constructed still lack in accuracy.² The combination of datasets dilutes the responsibility of each service provider from whom the datasets are obtained. Despite their complicity in guilt, the service providers can easily point fingers to one another, making it hard for the users to find a particular target to blame. And if a user inadvertently consented to revealing his or her identity regarding one particular service provider in the first place. Then the blame can be shifted to that user.

A Way Out

Upon review, I realized that it may not be productive to argue for the right to control data about ourselves in order to seek changes in the practices of service providers. I cannot avoid making a choice between convenience and privacy. The calculus partly depends on exactly how much convenience I would have to sacrifice, so for the past week I set out to install a personal cloud in order to get a rough frame of reference.

After two hours of preliminary research, I learned the kind of hardware I need to purchase: a single-board computer such as Raspberry Pi plus accessories such as heat sinks and a USB keyboard at a total of $86.92. Once I obtained a new Raspberry Pi, I spent another two hours in research to learn how to install the images of the operating system and how to communicate with the equipment using my Windows laptop. At that point I was finally in the position to follow a guide online on installing personal cloud on Raspberry Pi, which took me another four hours to complete. Finally, in order to make my cloud accessible outside of my local network, I spent another two hours setting up port forwarding and dynamic DNS. In sum, I spent 10 hours of my time and $86.92 of my dime to set up a workable personal cloud with 32GB of storage. In return, I gained the consolation that I can now store personal files without any privacy worries, conditional on my setting up everything properly without leaving any loopholes that hackers could easily take advantage of. I most definitely have left some loopholes in one way or another, but at least the hackers or the government will have to work retail. You can access my personal cloud here.

Set ALLOWTOPICCHANGE = TWikiAdminGroup?