There's clearly demand for discussing this topic, so let's do it here instead of the comments.

room topic changed to What info does Mma send back to WRI?: (no tags)

It seems that arriving to a good answer might require some joint effort, which can be cooridnated here.

Obviously it sends out strings for interpretation to Wolfram|Alpha

@halirutan it is on-topic, but if you read the chat, the answer "tcpdump does not capture any traffic if internet connectivity is selected "off" in the settings" is not satisfactory.

if the question is "does mma, without having root, send stuff which is not visible by simple methods such as tcpdump" then do you think this is appropriate to this site?

and yes I also don't think the wolfram sites are the place to ask this, obviously!

This appears to have come to a sudden halt.

well people are voting to reopen on the main site, so probably not

I wonder what kind of answer would be satisfactory though

This is probably a question that stackexchange cannot answer since they aren't Wolfram Research

If you want a formal statement about this, I would contact Wolfram Technical Support with this question (support@wolfram.com).

This isn't the normal kind of stuff they answer and there's a new release, so it'll be hard for them to answer probably. It might take a while.

AFAIK, the only information anyone at my company sees is what is sent when making a Wolfram|Alpha query

So things like what query you made. What your license and activation key are.

I believe it's documented under the WolframAlpha function what that all is.

So there's statistics done on what kind of queries are done for Wolfram|Alpha and I assume the data paclets and probably the natural language interface.

I mean when you click on something that sends a feedback to the company, such as a feedback report for the predictive interface, that obviously sends us an email with the information you see in the email (what the predictive interface gave you as a result and what version of Mathematica you are using).

I'm not sure if I could come up with an exhaustive list of such feedback mechanisms, but I believe it's always clear what is being sent with those.

@Searke If you read the comments, it seems that the point is what information is being sent not in the clear. Basically, the suspicion is essentially of foul play.

This is why it's asked here.

Oh...

I mean I never thought it was unclear.

@Searke What is not unclear?

Well there is functionality in Mathematica that sends feedback or sends things to Wolfram|Research

Those things are documented with what information they send generally

or they are things that bring up an email editor which fill out what is being sent

@Searke I am saying what I said because I suggested one can look at the output of tcpdumpin a comment, and apparently that would not necessarily be a sufficient answer

But yeah, if anyone feels like they want a formal statement from the company I can understand that.

Many people I know at the company value information privacy and are advocates for it.

I strongly doubt any statement from the company would be enough for this.

I would agree with that.

The question is seeking answers here though. And I'm afraid there's not much of an answer that can be given here beyond what you have already said.

Or me saying "I haven't seen us do anything evil."

That is what I thought too. But clearly others did not.

I think the only data collecting thing I've really ever seen was when we collected people's FB information with their consent via the Wolfram|Alpha app for that blog post.

I think at this point it's stayed open more as a matter of principle than for any practical purpose.

@Searke I agree with you, but it seems that the original poster wants some guarantees that Wolfram Research is not doing something nasty on purpose and doesn't try to hide any nasty behaviour from their users. IMO at this point the question becomes off topic on this site, as it's about techniques that can detect "nasty" behaviour, network communication, etc. It's not about Mathematica programming, nor about doing things with Mathematica.

There's an option to turn off internet access.

@sz

@Szabolcs yeah I agree.

It's a not a statement from WRI that he's looking for. It seems the question was posted with the assumption that WRI is not trustworthy.

@Szabolcs That's fine, the problem is that simple inspection also isn't enough. So it actually is a question on detecting hidden network activity. Which is interesting, actually.

I mean deliberately hidden

Yes. The reason I want them to contact WRI for a statement is because that's the best they can really get for an answer

Rumors can be hurtful and I don't want people to think that we don't care or value their data privacy.

@acl Right. But detecting hidden network activity is not something specific to Mathematica which I think makes it off topic. People disagree though, which is why it was reopened.

@Szabolcs Yes, that is why I voted to close it.

I don't think people kept it open because they expect an answer (especially once it turned out that we're talking abut actively hidden connections).

@Searke In some circles it is a given that WRI is evil (also Microsoft is evil, etc.). This attitude is what turns me back from using Sage. Every time I search for Sage solutions, I bump into this political nonsense, and I find it disgusting. Other communities around open source projects, like Python, R, etc. don't suffer from this bad mentality and there's no constant badmouthing.

WRI also has some disgusting marketing materials which gives me very bad feelings: wolfram.com/mathematica/compare-mathematica/…

but the community around Mathematica is not like this.

@Szabolcs I (and others) didn't reopen because we think WRI is evil... just that it is answerable to the extent discussed here in the comments. There has been a recent trend of punting everything to WC just because it mentions the company (only for it to languish there). Of course, it turned out that OP is simply paranoid and nothing can be done about it (and we look like fools for reopening it :D)

@rm-rf Hang on, who claimed you reopened because WRI is evil?

I didn't claim that, just to make it clear.

I am just referring to how popular WRI bashing is in some circles, and that is transferred to the product itself, and then eventually people using the product, which is what I don't like.

Anyway, I don't see that there is going to be an answer that is satisfactory without doing some tests and looking for (possibly hidden) network traffic.

@acl I didn't say anyone claimed that :) Just preemptively clarifying that because your statement and Szabolcs's were side by side and I read it together initially

I made this chatroom so there can be discussion about those test, and everything doesn't go in the comments. I am waiting for nathan to register an account so he can be given access to the chatroom and participate in the analysis if he wishes.

@Szabolcs I'm sure he considers SE evil as well and won't register an account :P

@Szabolcs Oh no, I understand completely.

@Szabolcs Registering an account is probably not going to happen!

I was raised on Linux and open source. I very much appreciate their part in the software ecosystem.

@rm-rf it's fine to be paranoid if you at least know what you're being paranoid about. here I doubt it

2. What information is sent to W|A precisely when internet access is not disabled?

I believe the answer to 2 is what is documented in the Wolfram|Alpha function

when you query Wolfram|Alpha, your license information is sent along as well to verify that your license is current for example.

Last year I wasted some time looking at the traffic captured by tcpdump over a few hours (of normal use). I could do that again for v10 (in principle)

Shouldn't the legal documents wolfram.com/legal/privacy/wolfram-mathematica.html and wolframalpha.com/privacypolicy.html give the answer.

@Searke I think the point here is, all this is independently verifiable via looking at traffic. That's not what he's asking about, apparently

I'll think I'll go back to rewriting my code to use Managed Library Expressions. Anyone using LibraryLink in v10 should look at this new feature.

or actually no, I couldn't repeat this for v10 because it just stopped connecting to the internet. I killed the library files, etc, nothing. doesn't connect from the command line either. oh well.

@Szabolcs oh that looks neat, I missed it

@acl It allows creating data structures on the C side and still automatically freeing them up when they're no longer in use by Mathematica. For example, TetGenLink had TetGenCreate and TetGenDelete. With this feature TetGenDelete no longer needs to be called explicitly (or at least this is possible to implement, I don't know if TetGenLink already does it).

@Szabolcs yes, reading about it now.

@acl Yeah. Also, on second thoughts, we should discourage "I heard from random X that WRI kills a kitten each time someone mentions undo -- Is this true?" type of speculative questions. If OP had used Wireshark or something equivalent and shown some concrete evidence of suspicious traffic (or linked to a blog post by someone reputable), then it would've been fine.

@rm-rf Absence of evidence would be OK too. Yes, that's what I tried to say in the comments, but no: this is on topic, should be open etc

Yeah I wanted to ask whether the person was a Mathworks sales rep

@mfvonh No look it's fine to ask this, I had the same question last year. But what he is actually asking about would take a security expert.

@acl I agree it's a legit question, but I also agree with @rm-rf that "so-and-so told me ... " without any sort of parameters for what the answer should be is not a good question

Especially because it sounds like rumor mongering

@mfvonh It would be OK without the so and so told me (in fact I asked and he did not say who is this "so-and-so").

@acl I agree - this site has the advantage of discussing things that might be too "critical" for the Wolfram community site, but in the context of this question we're stuck, it seems. I think I'll vote to close it again, since it could be said that it needs input from Wolfram experts and on the other hand nobody on this site seems to expect a real answer to be forthcoming.

This reminds me of a question that started "Mathematica's front-end is useless, shit, and does not deserve the name". Mr.W edited it to "I find the frontend lacking." :)

@Jens I could have written one on what one can see by inspecting traffic superficially, but mma stopped connecting to the internet. So, I can't.

Some of you might not be able to vote to close again (if you've already closed it), but we can do a show of hands and I'll close on behalf of all (or you could also edit it to make it proper)

@rm-rf I am not entitled to vote on such matters until I get another 25 rep :)

Just 25? brb ;)

@rm-rf OK, although I could still vote, I'll just show my hand in favor of closing right now and let you guys decide.

@rm-rf I feel like Mary Poppins with my rep umbrella :P

closing would be my suggestion, too

A statement like this one form WR could be enough webinstall.com/consumers/our-privacy-policy

@Searke Can you help fill out the missing details? mathematica.stackexchange.com/a/54936/12 The WolframAlpha function documentation doesn't actually say what information is being sent. Could also add more information on how WRI servers are contacted (paclet updates, documentation updates).

Since the question has so many upvotes, we might as well answer the reasonable part of the question, and ignore the part about intentionally hidden network traffic ...

@Searke Other interesting question you may be able to help answer: does e.g. Sunrise or Today trigger sending info to Wolfram servers? What info? Is any info sent on startup? How is $GeoLocation computed?

Here's the Mathematica Privacy Policy

> Certain functionality in Mathematica allows integration with Third-Party Sites (TPS), including but not limited to social networking sites such as Facebook. Should you utilize such functionality, we may collect personally identifiable information about you from the profile you have established at the TPS.

> By authorizing Mathematica to access your TPS profile, you are authorizing Wolfram Research to collect, store, and use in accordance with this Privacy Policy any and all information that your privacy settings at the TPS allow Mathematica to access through the TPS application programming interface (API).

@mfvonh Like this monster en.wikipedia.org/wiki/Universal_Church_of_the_Kingdom_of_God

Fromwikipedia

Does this mean that whenever I use functions like SocialMediaData to visualize my Facebook friends, WRI will store that information and may use it for its own purposes, such as creating that Facebook analytics blog post?

Well. I'd first see if Sunrise[] needs an internet connection

Block[{$AllowInternet = False}, Sunrise[]]

It does, so it's either querying a paclet server or wolfram alpha

in general this entails sending the things you see in the documentation for Wolfram|Alpha

like your license number for example

activation key

If they don't check out, then the query might fail

Or if you are calling Sunrise[] a billion times a second etc.

$GeoLocation depends on $GeoLocationSource

basically it's GeoIP

which is guessing your location from the IP address right? I've never actually questioned how that worked....

There's a mobile option for that since it's possible different devices might have options for identifying their locations.

If you want to see what Information is being sent in a WolframAlpha query, you can look at the URL which is generated by the WolframALpha function

WolframAlpha["2+2", "URL"]

What the function does is it generates this url and then uses it.

Does Mathematica send information when it is started? I don't think so.

I'm not sure why it would

I guess it could send some license info to a server to check that it's valid. Might help with piracy

@Searke I asked about that because I do not know how geolocation works and whether it requires sending one's IP address somewhere.

The way activation works is as documented. Although if I remember correctly it's possible that a site license might update somehow and I guess that would require the possibility that it might contact the sever looking for an update when started. ... I think?

@Szabolcs I really don't know much more than that really. So yes there is a process by which the IP address becomes a location. I guess it's possible that this might be a third party service.* shrugs shoulders*

From what I understand, it'd be kinda silly for us to hire an outside service. It's apparently not that hard. But If you want to know for certain, yeah you'd have to contact support

I think I should also note that this is for Mathematica

Doing evalutations in the programming cloud could be different

obviously since you're typing something that goes directly to our servers, everything you evaluate goes to our servers.

^Tautology

I don't know if they collect statistics beyond usage or not