# How to distort data in a clever way? [closed]

For example, I have some time series. How can I change my data, so it will not be obvious to understand what was original values?

Ideally transformation would allow to revert and reconstruct original data with some noise(or to save the relationship between time series values with no option to revert to original values)

I thought of: multiplying each value of series by some number, which will do shift and scale operation or I can compute some continuous function(like log), where argument is my time series data.

Nevertheless, I believe there should be more advanced method. Thanks!

Update: After careful considerations, and answer of @D.W. below and trying to transform data, I found that I spend more time on this than on main problem I have.

Initial problem was twofold:

1) to send real data to home computer to play at home any time I want.

2) to send real data to home computer, transform it a bit, to share with people in the web, to get some advise on my main problem.

After thinking, stressing, askying, googling I decided to drop the idea to do anything. Few good links to people who have the same problem and are intended to transfer, transform data:

For problem 1) I believe stream cipher is a an option to read more about ( as it transforms your data to a random string). Result is bad, as if one will look at your working computer logs, you are busted.

Regards.

• Welcome to the site! What's the goal; to encrypt the signal? If so use encryption. Otherwise can't the adversary reconstruct the signal too?
– Emre
Jul 13, 2017 at 20:32
• Dear @Emre, kind of. I just have noble desire to play with the data at home, while university does not allow to upload it to my computer ( I have no back thought). I have several time series, and I have to preserve their relationship after transformation. Shortly, how can I distort data so that nobody guesses what is that? Is encryption the only way we can think of? Jul 14, 2017 at 7:42
• What's your purpose here? Jul 15, 2017 at 9:39
• Dear @edouard, I added update Jul 16, 2017 at 13:03
• cs.sfu.ca/~jpei/publications/… Sep 17, 2017 at 0:17

The short answer is: don't do that; anything you come up with on your own is likely to leak information in subtle ways.

The longer answer is: talk to your data protection authority. Only use a method if they consider it acceptable. (Hint: it's very likely that the answer you will get back is "no, don't do that".)

The technical details are: there is an entire body of work on how to anonymize or de-identify data, so it can be revealed in a way that can't be linked to the people whom the data is about. However, there's a long history of attempts to do this that failed; where it was actually possible to re-identify the people, violating their privacy. See the following research paper:

Myths and Fallacies of "Personally Identifiable Information". Arvind Narayanan and Vitaly Shmatikov. Communications of the ACM, 53:6, June 2010.

If you wanted to consider a scheme to do what you want, you'd need to tell us a lot about the nature of the data, what it reflects, what the privacy interests are, how the data is distributed, what other external sources might be available that might enable re-identification, and many other details. Since you haven't provided any of that information, your specific question isn't answerable in its current form.

But ultimately, I hope I've convinced that what you're trying to do is a bad idea. Don't endanger people's data by going "cowboy" with some homebrew scheme to obfuscate the data; instead, talk to your data protection authority about what their requirements are.

• Dear @D.W., I am following approach offered by you: talk to data protection authority. I think I will drop the idea of moving information out. Jul 16, 2017 at 12:37
• thank you again, probably you saved me, even I had no bad intentions. Jul 16, 2017 at 12:44