Hacker News Reader

Comment by ewzimm | Hacker News Reader

This would be too big an oversight for a company trying to get more business adoption. According to Microsoft, the data is anonymized and run through a virtual shredder before being sent. Here's their statement on it:

“Some of this data is stored on your device and some is sent to Microsoft to help improve these services. Data sent to Microsoft for product improvement is put through rigorous, multi-pass scrubs to remove sensitive or identifiable fields (such as email addresses, passwords and alphanumeric data) and strings are chopped into very small bits and stripped of sequence data to prevent the information from being identified or put back together.”

http://blog.laptopmag.com/windows-10-privacy-issues-exaggera...

It amazes me that this information isn't easier to find and I end up being the one defending Microsoft. I definitely recommend using Free Software only if you want to be absolutely sure of confidentiality.

View on HN

Do I understand this correctly that they are sending back everything you type, but they try to scan the data for things like e-mail addresses or passwords and remove those bits first, and then send the data back in pieces so it can't be reassembled?

If so, that sounds unbelievably error prone. What if your password is of the "correct horse battery staple" form, rather than the standard unreadable scramble? It seems impossible to reliably distinguish that from other text. Does that mean that Microsoft then has your password? Even anonymized, that's a massive problem.

It's not everything you type, just bits and pieces stripped of sequence, and most of what's saved is on your device only. They specify that what they send is only "some," and logically, everything typed into the billion devices they're shooting for would be too much data for even Microsoft to handle.

Passwords and the like are probably identified by the type of field they're entered into, as Windows forms and websites have explicit identifiers for passwords, and it would be easy not to record this. I think this analysis is more for things like analyzing the frequency of terms and which characters tend to follow other characters, but I don't have much more information, and I agree that there needs to be more shared about how this works.

So, here's a question then, assuming that's true, why send it? If its been so thoroughly anonymized that nothing useful can be gleaned from it, why bother sending it in the first place? If it is possible to derive useful info from it, that seems to directly contradict the idea that no sensitive information is going to be exposed (there being no realistic way to programmatically identity only sensitive information without some very specific domain knowledge). You can't have it both ways, either the info isn't useful in which case they have reason to transmit it, or else it's exposing potentially sensitive information and they shouldn't be transmitting it.

I can only speculate, but I would guess it goes something like this:

You type "the quick brown fox jumps over the lazy dog." You continue typing other stuff. 30 minutes later, there's a queue of random slices like: { "lazy dog", "chocolate donut", "brown fox" }. That gets encrypted and sent into a big pile of other data for a machine learning algorithm to develop better typing prediction.

I would like to know if that's the gist of it or if there are other things happening. Hopefully there will at least be a developer blog or something like that to better explain it.

That's what it may be sending today. What about next month after one of the forced automatic updates? If you authorize someone to take your data, that doesn't mean they have to do it right away.

> only "some"

99% is still "some". Regardless, since when is "some" keyloging acceptable?

> too much data

That depends on a lot of factors. Text is small and trivial compared to most stuff on the internet these days. How many kilobytes per day of text does the median windows user even type?

It would be even easier for MS to filter out stuff like video games (DirectInput?) to reduce the size.

> I think

Guessing is useless. What matters is the data going over the wire and what authorized MS to do in the future.

> there needs to be more shared about how this works

No, there needs to be explicit informed consent form the user.

I'm just trying to objectively interpret a really small amount of data here, not make a justification for or against any particular policy. You're right that it's important to have informed consent about this kind of thing. That applies to Google, Facebook, or Amazon too. It does seem like it's pretty obvious here when there are so many statements about how Cortana gets to know you, your speech, your schedules, your flights, your interests, etc., and there's a delete button for everything too.

Most people probably like all that, but I would like to see more robust privacy settings to easily disable tracking whenever it's desired.

> That's what it may be sending today. What about next month after one of the forced automatic updates?

You mean, like this?

# apt-get update && apt-get upgrade

Granted, running that isn't 'forced', but given that OSS users don't _actually_ analyze the code in their software, it's basically the same thing.

There's always a chance of something being slipped in later, no matter what OS you're on. Rampant paranoia and conspiracy theorizing does no one any good.

The "forced" you so easily dismiss is entirely the point. In case you haven't noticed, people choose when (and if) they upgrade, for a variety of complex reasons. One of those reasons is to wait and see if the early adopters complain of any issues.

The problem of inspecting the code itself is a separate issue entirely, though at least in the area of Free Software that inspection is possiblea, and reproducible (deterministic) will simply the problem. Good luck doing the same with Windows (or any other closed software).

> Rampant paranoia and conspiracy theorizing

Now you're simply resorting to either insults (and/or wilful ignorance).

> The "forced" you so easily dismiss is entirely the point.

I think it's you that missed my point.

> people choose when (and if) they upgrade, for a variety of complex reasons

If they're running Gentoo, or maybe a small group of other distros, sure. Otherwise, people (on all operating systems) don't really do a lot of choosing at all. If updates are available, they generally run them. The only picking and choosing I've ever seen from the vast majority of people (including technical folks) is if they don't have the bandwidth/time to update a particularly large piece of software at the moment. Of course, they'll update it later when they are able.

> The problem of inspecting the code itself is a separate issue entirely, though at least in the area of Free Software that inspection is possible

Yes, it's possible, but it isn't done in practice, which is why I said it. That's the reason I said it's 'basically the same thing'. Whether you receive updates automatically or choose when to download them, you are still accepting what the distribution method gives you. There are FAR too many software packages out there for the average person to work out the logistics of reading accounts from early adopters of every single point release.

> Now you're simply resorting to either insults (and/or wilful ignorance).

It's not an insult. It's a factual statement. You claimed in your comment that a multi-billion dollar corporation who has a vested interest in keeping the good will of their customers (some of whom are corporate, some government, but the vast majority are consumers who look to them for help) would willfully send compromised software through forced updates in the future. If that isn't a conspiracy theory, I'd love to know what is.

> The only picking and choosing I've ever seen

Apparently you need to get out more, because you seem to have missed that "not upgrading" is so incredibly common that some major software packages (web browsers) and at least one operating systems have decided to force upgrades.

Also, do you really believe that all the IT departments in the world are running current versions of anything? There are entire industries that run "last year's" versions, who only upgrade after extensive testing.

> I'd love to know what is

Did you even read the thread you're replying to? You're knee-jerk application of the paranoid-style[1] ignores the fact that "It's in the privacy policy." There is not secret. MS is telling everyone they will send forced updates (which have already cause stability issues for some people). As for MS keeping "the good will of their customers", again, did your read this thread? Windows 10 has already forced various professionals to drop windows.

Also, "compromised" is your term, not mine. I'm sure any future update they push out will be buzzword compliant. The point is we don't know what future updates will bring, but you agree to run those future versions anyway.

Meh. I have better things to do than waste any more time on your distractions. If you want to be the apparatchik that refuses to believe MS would do anything malicious, that's your business.

[1] http://harpers.org/archive/1964/11/the-paranoid-style-in-ame...

> "not upgrading" is so incredibly common

Honestly it does not look like you wasted much time at all, you seem to have missed his point again.

The trouble is that not every place you type a password will necessarily have something identifying it to the OS as a password field. For example, what if you're entering a sudo password into an ssh client, or saving a password into a text file (saved on an encrypted drive, surely) using a text editor?

I'd like to know too. Maybe they compare typed data to a dictionary and throw away everything that doesn't match so that every slice contains only dictionary words that couldn't be identified as passwords. That's just speculation of course.

It's hard to see the utility in a bunch of individual unassociated dictionary words being sent back. Maybe there is some that I just don't know about. However, my money is on them being more confident in their scrubbing than is warranted. There have been so many cases in the past where companies have scrubbed and anonymized data, only to find that they missed something critical.

Even the frequency of words used might be valuable for typing predicion, but I would guess they would probably group a few words together. I really would love to know how it works.

Same here. If it has holes, it's important to know about them. If it doesn't, then it would be fascinating to understand how it works.

They made a statement to a magazine. Where is that statement on their website, and where is the information that discloses what they are sending?

I'd like to see precisely how they anonymize the data, and I'd also like to know what data it is.

Because companies haven't failed to anonymize data before with the general public being left to pick up the tab.

If Microsoft was a known moral bastion who had perfect security might not be a problem. But it isn't and it doesn't, so there is only one response: NO!

(OK, there are more responses, but I'm not sure if they are allowed with HN's code of conduct.)

> I definitely recommend using Free Software only if you want to be absolutely sure of confidentiality.

So, has anybody ever audited build farms for Debian, Ubuntu, etc? Have we ever seen a breakdown of their income sources? How do you know that the binaries you install are built from the sources you can browse or download from the net?

Do you compile everything yourself?

Debian and others realize that this is a problem, and you absolutely shouldn't trust anyone at their word. There's an ongoing effort at Debian to make all builds reproducible so that you can verify that they were built from the sources they provide for inspection. This should be completed soon.

For the situation I'm talking about where absolute confidentiality is required, the right way to do it would absolutely be auditing all code and compiling from source, assuming you're working in an organization that has that kind of capacity.

But for most cases, just dealing with an organization you trust is enough, whether they provide source or not.

> "There's an ongoing effort at Debian to make all builds reproducible so that you can verify that they were built from the sources they provide for inspection. This should be completed soon."

Ah, good to know. Do you have any links to articles or discussions about this? Would like to learn more.

You can trust Microsoft. Don't worry about it. They'd never do anything to compromise your privacy. Just ask them.

/LOL.

While it is true that most end users of Free software do not audit or compile their own binaries, the situations aren't even remotely comparable.

FWIW, I believe they try to do these things however, I'm not sure how its technically possible.

Various security measures rename sensitive fields into things that are basically random strings that are stored in a session. [e.g. Fields to change your password? They are set to something like 2ajefklaj23324rk]

There is also the issue of people who write erotica for fun. Or doctors writing down notes for patients.

Here are some references to explain how data is anonymized:

http://www.caida.org/data/anonymization/index.xml

How do you think Apple gets their spell-check data, or improves Siri's voice recognition?

In any event, during the installation process, you do see a screen where you can turn all this off. It's not off by default, but viewing these settings was part of my Windows 10 installation experience.

I thought the options to turn off defaults were pretty clear during the installation process. Requests to make options more explicit and visual are reasonable, but it bothers me that Mozilla and others are acting like "choose my default programs" or "custom settings" are too obscure for anyone to understand. You also have the option to delete your history.

Turning those things off still doesn't prevent Windows 10 from sending tracking and usage data to Microsoft though, even when you're only using a Local Account instead of a Microsoft Account.

I don't know, how about using a dictionary and doing research?

Otherwise wouldn't every commonly misspelled word show up incorrectly in spellcheck?

There is quite a bit of work in the field of speech recognition, too, and it's entirely possible to create a functional model without recording every user's voice and storing that information on a server. Even if they require speech samples, or samples from different languages, they should at least pay someone for that information.

I was disappointed when I learned that the voice recognition feature in Siri wasn't local to the phone.

I can confirm: I bought a Surface Pro 3 the other day, and when I first booted it up, the setup screens explicitly asked me which privacy settings I'd like to enable.

Do you know if there is any information sent to Microsoft by default for which they did not offer you a screen widget to click that says "off"?

Do you know what actual data is, and is not, sent to Microsoft as you use your device?

Even if you do, will you still know once Microsoft pushes a mandatory update and/or changes its 12,000 word terms of service?

> In any event, during the installation process, you do see a screen where you can turn all this off. It's not off by default, but viewing these settings was part of my Windows 10 installation experience.

The problem is that Windows 10 does not respect you having this turned off. It will send away your data, even if you turned off all the switches.

Don't you see this as a huge problem?

Reply on news.ycombinator.com