Dropbox Is Probably Not Stealing Your Files | Hacker News Reader

Dropbox Is Probably Not Stealing Your Files

one.darrenpmeyer.com

140 pointstalon8899 comments

This is a fine post, but all I can think about this situation is "or, you could just reverse the Dropbox client and find out for sure".

Speculation about Dropbox stealing files seems premised on the idea that you can't know what the client is doing. But that's not even close to true. People reverse much, much harder targets than Dropbox for fun. If any version of Dropbox published to its user base ever did anything like this, we'll all know soon enough.

I would be very surprised if a workplace name like Dropbox has never been reverse engineered by a bored hacker on a lazy weekend.

Surprised and disappointed.

I reversed it back when it was version 1.1.something, it was basically all compiled Python modules with custom encrypted code objects and non-standard opcode mappings for the bytecode.

Quite interesting to see how it worked, and useful to get the key for the encrypted logs, to see it what it actually did while running. Back then you could intercept the https connections as well as they hadn't pinned the certificates yet, to get an even fuller picture.

There was nothing obviously nefarious going on back then, but that was quite a few years ago of course.

So what you're saying is that, if I wanted to launch a nefarious file-stealing Dropbox-like application, I should first launch the non-nefarious version, and then when it gets up to 3.6 or so, turn evil?

Yes. That's exactly the right time to turn evil.

https://xkcd.com/792/

I used gdb to talk the Dropbox client into opening a repl.

For examples see pyrasite, code.interact, etc.

If you specifically want to know what files Dropbox reads there are easy ways to observe this, like strace.

Previous (successful) attempts at reverse engineering: https://www.usenix.org/conference/woot13/workshop-program/pr...

What happens here is that Windows Explorer asks Dropbox whether it should display a green or blue icon for a file. It does this for all files, including those outside of the Dropbox folder.

Full analysis here: https://news.ycombinator.com/item?id=9139657

I couldn't be fussed to reverse it again, so instead used Rohitab Batra's excellent API Monitor [1] to figure out what's really going on.

Firstly to address files being accessed outside of Dropbox - this is true, but literally all it does is read the file attributes: https://i.imgur.com/TADvHp1.png. Moving up the call stack and disassembling the calling function, we can see that it's part of the Python runtime: https://i.imgur.com/1TBong4.png (presumably python27_lockdown.dll is Dropbox's custom hardened copy). A bit later on it does a comparison to ".bat", which identifies it as the function win32_stat() in Modules/posixmodule.c - the ensuing behaviour of this function corresponds to QueryBasicInformation as shown on the original author's Process Monitor dump. Why the Dropbox client calls stat() on files outside of the Dropbox folder (but on the same drive) is not clear, but, as the article above also mentions, that is all it does, so no problem there.

Secondly, the original author also posted evidence of Dropbox accessing various shell folders [2] - Desktop, Documents, Music, Pictures, and so on. This is true but again it's a side effect of an innocent function call, this time SHGetFolderPathW(): https://i.imgur.com/uXN31BI.png. It's actually SHELL32.DLL that is responsible for opening the folder and querying its attributes, not the Dropbox client: https://i.imgur.com/YCyTwNe.png.

Without reversing the entire program we can't say for sure that Dropbox isn't siphoning out data in some other sneakier way, but the accusations of data theft from these file events are simply not true.

[1] https://www.rohitab.com/apimonitor

[2] https://pbs.twimg.com/media/B_Kv4i3U8AEZLLt.png:large

Those were my exact thoughts.

Even without having to go open IDA, I'm sure windows has enough system monitoring tools that you should be able to tell what Dropbox actually reads outside of its own data, if anything.

"This is, by necessity, a system-wide process."

This is, by design, a fucking huge defect of the underlying system calls.

Give the OS a list of folders to watch. Dropbox should not even get a callback for a file it's not is supposed to watch.

See: inotify in Linux

inotify is by far the best of the filesystem notification APIs. It's wildly reasonable.

You can constrain filesystem callbacks on Windows, I have no idea why Dropbox doesn't.

Exactly - how much information can be inferred from the file and path alone - in a lot of my cases (for very personal files), quite a lot.

If Dropbox gets to interpose itself in every file update systemwide, that sounds like it's getting too much of my information. Encrypted/Stego drives don't matter - once you open it up to the system, Dropbox sees when you touch it.

All I can gather from this discussion is that I'm glad I'm not using Windows and Dropbox together.

Quote:

"A simple protocol can give us an idea of whether data is being sent to Dropbox:

1. Create a large-ish file (1MB) outside of the Dropbox folder

2. Monitor the network usage of the Dropbox application to see if it sends enough data that it could be that file

3. Repeat with many different files, etc.

Doing exactly that, Dropbox only sent a few hundred KB after “accessing” the target file. Seems unlikely that Dropbox is uploading files outside your Dropbox folder."

This test approach has a problem. A more realistic test would be to place a well-compressed file, one that by definition cannot be made smaller, on Dropbox and see what the system traffic size is for that file. For an optimally compressed file, if the system is reading the entire file, the read size will more or less equal the file size.

You’re still missing the possibility of DB uploading a hash vs standard file compression. Even if they never upload the full file plenty of people would love to know if any of your files was on some list. (Classified information, piracy, etc.)

If you read the discussion near the end about why a few hundred KBs, you'll see that the first test file I used was a JPEG embedded in a Word doc (the latter because the original claim used Word files), which should be very difficult to compress.

To be fair, I added this section after several people pointed this out, so you may not have seen it.

Well, dropbox could just listen for fs events inside the DropBox folder; and it should, from a performance perspective as well as from a privacy point of view. And then "sends a few 100 kByte"? I hope this is a typo; if not, I would like to know what these are. (also: the OP's largish file (1MB) could easily fit into "a few 100kByte" after compression)

Actually, it cannot for the shell extension for the checkboxes over synced files. It needs to look at every file seen with Explorer to do that.

Really? The shell extension could just ask the dropbox daemon if something is known about file x/y/z, and the daemon would answer "no" if the file is outside its folder. No need to look into the file for that. Or did I miss something?

If you look at the analysis posted elsewhere, Dropbox only queries the long-name of the file since apparently Explorer can still send it a short-name version of the file (e.g. "C:\PROGRA~2" instead of "C:\Program Files").

Well, dropbox could just listen for fs events inside the DropBox folder...

On Linux at least [1], this is exactly what the Dropbox client does. It only registers inotify watchers on the $HOME/Dropbox directory and subdirectories. To verify:

    strace -f -e trace=inotify_add_watch dropboxd

You could also strace open/stat/read/write syscalls to verify that, aside from shared libraries and the like, the Dropbox Linux client doesn't access files outside of your Dropbox directory.

Other OSes have different file monitoring capabilities though. Anyone up on file monitoring on Windows / OS X? Is directory-specific monitoring possible?

[1] https://www.dropbox.com/install?os=lnx

> Other OSes have different file monitoring capabilities though. Anyone up on file monitoring on Windows / OS X? Is directory-specific monitoring possible?

yes

But ReadDirectoryChangesW notoriously misses updates. Also it would scale horribly to large amounts of files.

NTFS has a feature called Change Journals where you can view a volume as a stream of changes.

OS X definitely does, it's quite simple and reliable.

The cycle of security is this:

1. Security experts see a security hole and note that it could only be used by a widely trusted company or government.

2. People note that it's also possible that it's not happening, and claim that the widely trusted company or government would never use the security hole.

3. It is discovered that the widely trusted company or government has been using the security hole.

"Touches files it doesn't have to" is not a security hole.

"Users install program that touches files the users might not want it to" is a security hole. The unknown is whether Dropbox comes with malware which takes advantage of that security hole.

A hidden function or update could enable malicious behavior on all files whether or not it had preexisting behavior of touching all files. Only in certain detailed permission structures would preexisting behavior matter.

If a hidden function enabled malicious behavior, causing it to touch all files, the hidden function would very quickly cease to be hidden.

Are you seriously arguing that it's okay for Dropbox to touch files you didn't give it permission to touch? This is ridiculous.

>If a hidden function enabled malicious behavior, causing it to touch all files, the hidden function would very quickly cease to be hidden.

I'm not sure where you're going with this. Yes, a security hole would become much more visible after it was exploited. That doesn't imply that anything visibly weird Dropbox does is a security hole.

The only notable flaw in security here is that it's a program on a normal OS outside a sandbox. This is a huge flaw but it applies to most programs.

>Are you seriously arguing that it's okay for Dropbox to touch files you didn't give it permission to touch? This is ridiculous.

I am. Touching files does not mean taking information from files. And between the explorer extension and the way file monitoring works on windows it's going to be fed a list of your files no matter what.

Security holes are a subcategory of "things a program can do, but shouldn't be able to do". They are described entirely in terms of potential behavior, not current behavior.

Okay, if you want to be pedantic about the meaning of the words "security hole" instead of addressing the actual concerns people have with Dropbox, then we can just call Dropbox "potential malware" and be done with it. Does that address your terminology concerns? Can we move on to talking about the important stuff now?

Perhaps it's semantics, but you're right - it's not a security hole.

It's worse.

Exceeding authorization is exploiting an existing hole.

Read their TOS and compare it to Google Drive's TOS.

The insane rights the Google TOS grants to Google are why it costs ~ 1/2 as much.

It is also an indicator that Dropbox is less shady. They don't grant themselves rights to do anything with your data outside of the normal things you need them to do to offer the dropbox service for your use.

Unlike Google, which could for instance, use your personal photos of your kid eating ice cream to try to sell you ice cream via road side LED billboards.

I think you're misunderstanding the TOS. I do not think Google is subsidizing their storage costs by abusing your private photos.

They might not be doing it (and probably aren't), but they do have the right.

Interesting how I'm getting downvoted for stating this. The language does come awfully close to explicitly stating just the thing:

"We collect information to provide better services to all of our users – from figuring out basic stuff like which language you speak, to more complex things like which ads you’ll find most useful"

even explained thus:

"For example, if you frequently visit websites and blogs about gardening, you may see ads related to gardening as you browse the web."

and they say that they base it on:

"Information we get from your use of our services."

Again, not explicitly named, but not forbidden either by their own privacy policy.

I think you aren't actually reading it. I do not think Dropbox is less good at running storage than Google or Amazon, yet Google is 1/2 the cost.

The reality here is that the cost to operate doesn't matter to Google. They can keep subsidizing Drive with revenue from Search Ads long after they have pushed Dropbox out of business, if they choose to do so.

This is why Dropbox (smartly) does not want to compete on price -- instead they want to compete on quality. There's no way to be cheaper than Google, Microsoft, Apple and Amazon in the long run, because they each have other businesses that are licenses to print money. If one or more of them decide to heavily subsidize their online storage product, they can outlast you.

However, it turns out that all of the money in the world can't magically make a great product. You still have to actually do the hard work.

We all know that the reason Google is running face recognition and photo recognition software on your all your files is to cross reference it with databases of paedophilia and terrorism !

Why the downvotes?

http://www.telegraph.co.uk/technology/news/11012008/Paedophi...

I'm aware of nothing in Google's terms which isn't in the terms of every other user-generated-content service that has lawyers on staff.

In order to make copies (distributed storage, network traffic, etc.) and show copies to people -- including you, or anyone you choose to share a file or post with -- they basically have to have a license from you, or they're potentially on the wrong side of copyright law.

Then you need to read the damned TOS side by side with dropbox. They are significantly different in scope and the rights allowed.

You claim that Google could use personal photos of my kid eating ice cream to sell advertising? This seems directly against their terms of service found at http://www.google.com/policies/terms/.

> Some of our Services allow you to upload, submit, store, send or receive content. You retain ownership of any intellectual property rights that you hold in that content. In short, what belongs to you stays yours.

Do you have any proof of this ever happening? Do you have any legal case that support your claim? Can you please point to the text in their TOS that leads you to believe this?

You need to actually read the TOS for dropbox. It says they can use your stuff for any service they offer now or in the future.

I don't need proof. You need to go read the TOS I mentioned already.

I think you may need to re-read the comments you yourself wrote and replied to.

Here ya go.

"When you upload, submit, store, send or receive content to or through our Services, you give Google (and those we work with) a worldwide license to use, host, store, reproduce, modify, create derivative works (such as those resulting from translations, adaptations or other changes we make so that your content works better with our Services), communicate, publish, publicly perform, publicly display and distribute such content . The rights you grant in this license are for the limited purpose of operating, promoting, and improving our Services, and to develop new ones. This license continues even if you stop using our Services (for example, for a business listing you have added to Google Maps). "

They changed the TOS to rephrase, but didn't remove the infringing portion. It used to say any service offered now or in the future, and now it says "to make new services."

The fact is, my original statement stands.

They can quite literally use your content for any reason because they can use it to develop new services (such as, say, personalized road side advertising.)

There is nothing legally stopping them from doing it, and if you have been paying attention to the issues highlighted by Snowden and others, there is little backlash to them as a company for doing very evil things such as leaving inter-datacenter communication unencrypted allowing the NSA and others to snoop upon Gmail and all other google services as the data replicates across locations.

Good times talking to you. Appreciate the down votes.

Once again, make sure you actually read the damned thing in question before even replying to a comment about it. You ignorant, uninformed points are worthless.

Also, here is the dropbox TOS: " Your Stuff is yours. These Terms don't give us any rights to Your Stuff except for the limited rights that enable us to offer the Services.

We need your permission to do things like hosting Your Stuff, backing it up, and sharing it when you ask us to. Our Services also provide you with features like photo thumbnails, document previews, email organization, easy sorting, editing, sharing and searching. These and other features may require our systems to access, store and scan Your Stuff."

REMARKABLY less ambiguous, and in fact enumerates the Services they offer. They literally only want to host your data. Google wants to data mine your data, even if you delete the data and your google account, google wants to keep using your data, forever.

This isn't the same thing.

" 1 - Create a large-ish file (1MB) outside of the Dropbox folder

2 - Monitor the network usage of the Dropbox application to see if it sends enough data that it could be that file

I can not really say if Dropbox steals them or not. But if i were a Dropbox engineer and want to know about those newly created files, i wouldn't want to send the whole file to server at all.

- Send file name with its extension

- Send file size

Compare these to dropbox's blacklist file (imagination only) in another server. If there are any matches, mark user as "whateveryouwant"

As long as there is a network activity when a new file created, it is and will always be suspicious to its users.

And they don't have to steal the whole file at once, why not chunk it out and do bits at a time over a few days. This is tinfoil hat territory, but they'd still get the file and network traffic wouldnt be overly suspicious.

You don't need to retrieve the file if you already have it.

As an aside, I've been using Sparkleshare (built on GIT and SSH) lately. It's pretty good, and sucks up less battery on my MBP than the Dropbox client (maybe because it's not watching every file in the system!) And not only is it open source, but you can see a log of all the git commands and fix things manually if necessary.

My only lament is that it doesn't work that well over the intermittent connections. It'd be neat to have something robust like mosh (https://mosh.mit.edu) for file sync.

There is also unison; it is not automatic (custom inotify script is required to mimic Dropbox, and it will force manual intervention upon conflicts) nor stores history, but handles large blobs, many clients and crappy connections quite well.

The original article is titled "Dropbox Is Probably Not Stealing All Your Files". From bandwidth consumption, you can tell it's not stealing all of them. Whether it does so selectively, on command from the mothership, is another matter.

"The Dropbox application uses a filesystem monitor to detect when changes are made by monitoring filesystem write events. This is, by necessity, a system-wide process. So DLP alerting that Dropbox is “acccessing” a new file shouldn’t be surprising."

I think this SHOULD be surprising to any competent software engineers. That isn't how the file system watcher works.

About every form of backup software will use change journals to identify what to backup and how it changed.

Change journals are streams that are per volume (so to monitor some directory in C:\ i have to monitor the C:\ change stream).

It's just how NTFS works. It's shocking that this was allowed to reach this kind of publicity because it's just a guy attaching a diagnostic tool to a system where he doesn't know whats happening and then proceeds to freak.

Software like this will have plenty of file access for metadata, not only on the backed up files.

And this is where having the API actually support, say, monitoring only items in a single directory would be good.

Well there is, but it's unreliable in certain edge cases. At least not reliable enough for a backup solution.

In which case the API is broken and should be fixed.

Also, what edge cases?

For example you have to allocate a buffer to hold the information you receive. If that's too small, you miss stuff. There is also a lot of intricacy with permissions and you can mess up a lot when multithreading without knowing how to properly interact with the OS then.

Wouldn't that be worse, if anything, for the whole-drive case?

This article makes me irrationally annoyed by how lazy the author was. I was able to produce a test in under 5 minutes that disproves the article's core assumption:

> The Dropbox application uses a filesystem monitor to detect when changes are made by monitoring filesystem write events. This is, by necessity, a system-wide process. So DLP alerting that Dropbox is “acccessing” a new file shouldn’t be surprising.

THAT IS NOT HOW THAT WORKS!

Sorry, I am calm now. As someone who has spent quite a lot of time using Windows' File System Watcher functionality, I know that that is nonsense. Windows monitoring/watching is conducted at the kernel, when an IO operation occurs that hits a registered monitor it fires off an event (windows message) to that process to let it know, the process itself never accesses that file directly.

But just test it for yourself.

1) Download Process Monitor [0]

2) Start Process Monitor, turn off Registry, Network, Profiling, and Process events.

3) Set the include (included processes to monitor) to [whatever executable you build]

4) Build this (see examples section) [1] in C#/VB.net and run it

5) Set the process name in #4 in the include in #3

6) Write to a file in C:\ (that's the default in the example program/source)

7) You should see some Console.WriteLine() output indicating the file watcher is working. If not run as administrator.

8) There you go. As you can see, no direct file accesses to the file. The monitor events are fired as you can see, but the file remains untouched directly by your program.

The author could have done this. Why didn't they? It isn't like I had to even write one line of code or have some kind of specialist knowledge of low level kernel functionality...

PS - I don't know/care if DropBox is stealing your stuff. I just wish the article's author had at least fact-checked before they claimed that "that is how this works!!!" when in reality that is untrue. That is how it works for Anti-Virus because AV scans within files to see contents, it isn't how it works for most processes which just use the file watcher functionality. If DropBox chooses to look inside files, then why? There is no need for that.

PPS - If DropBox do have a system wide file watcher, that is just lazy. It will reduce system performance, and they could have just as easily set it up to point just to folders DropBox is configured to watch.

[0] https://technet.microsoft.com/en-us/sysinternals/bb896645

[1] https://msdn.microsoft.com/en-us/library/system.io.filesyste...

It's very unlikely that dropbox would upload every changed file from your computer, that would not go unnoticed.

A desirable capability would be on-demand upload or download of any file on the clients system. For that you would need the entire filetree+checksums so, imo, that's what it's syncing.

Dropbox has too much to lose and not enough to gain by stealing your files. The accusation borders on paranoia. That said, Dropbox is a closed system, and I always trust open systems more.

The tin foil hat is strong with this thread. People should read up on Windows Explorer shell extensions before making comments. It's like saying regexing email addresses to check for valid input is the same as stealing emails addresses.

If dropbox get to send metadata about files outside of the folder could be damning enough.

Hey you have in your downloads folder 3 new files per day whose file names hints they were send by FB user X. I could make an educated guess about their content.

Reminder that Dropbox was mentioned, by name, in NSA documents released two years ago as the next target they intended to subvert. Also reminder that not long after that they named Condoleeza Rice, celebrated apologist for warrantless wiretapping during the Bush administration, to their board of directors.

Yes, it's possible they named her to their board in good faith, and it's possible they also resisted the NSA somehow, where Google and Microsoft and Yahoo and countless others failed. But, do you consider it likely enough to bet your privacy on it? It seems to me you would be foolish to do so.

The only other excuse for it I can think of is that it's so obviously corrupt that it proves they aren't corrupt after all - that no one could be that stupid. I reject such meta-reasoning. They are simply corrupt.

If you are an ordinary guy then even if Dropbox steals a file it won't matter much. If you are a government, Airbus, Snowden, Aselsan, or Comodo you should not install Dropbox even if you trust Dropbox.

Pure speculation follows.

I wonder if they could be calculating hashes of files and sending them off? That would be useful for automated exfiltration and targeting.

For example:

    1. Calculate the SHA-256 hashes for files in places of interest.
    2. Report the hashes upstream.
    3. Hey, this file matches one that the FBI/NSA is looking for via NSL.
    4. Download more stuff. Also identify the person and their location.
    5. Send agents/drones after them.

This is unlikely, but still in the realm of possibility. It's also untestable without more information. (Packet captures from the DLP device would be far more helpful in determining if anything of the sort is happening.)

Why would a TLA look for a file with a known (hashable) content? Is that common? Is that to be seen as looking for "contraband", i.e. if you have a certain file, for instance some known part of a rookit, you might be an evil attacking hacker? I don't quite follow.

Also, the typo "Drobpox" was fun, that's a good alias when feeling suspicious. :)

Tracking spread of leaked documents?

Or things like child pornography. It isn't all about state secrets, y'know.

To calculate the hash, it needs to read the whole file, which this post claims it isn't doing.

I appear to have overlooked this detail. Good catch! :)

Did the author actually verify this with strace (or the mac/windows equivalent)?

It sounds like he guessed this based on I/O activity of the process. It could be enough to hash the beginning of the files, and compare the rest if a match is found in the database.

Dropbox doesn't read the file content. There is also no proof that Dropbox directly accesses those files.

Not really, one could get a unique enough hash by reading the first lets say 10,000 bytes of each file, and it would be faster than hashing the whole file

edit: here i was bored enough > http://pastebin.com/NJEvnG1d

I wrote something that was hashing audiobook files that was taking forever, so I tried using the first N bytes (likely much more than 10kB), but soon found that for any given audiobook, each chapter's MP3 had a large identical header on the front end - I imagine that it was a cover image embedded in the metadata.

I think in the end I just started taking the data from the end of the file, but if you're going with subsets, it's probably better to use a pseudo-randomly selected subset rather than a sequential subset. It doesn't have to be a different pseudo-random subset for each file, but I imagine there's an ideal noise profile in the sampling (maybe white noise is best).

Maybe:

    File > 12 KB: First 4 KB, last 4 KB, middle 4 KB
    File <= 12 KB: Just hash the damn file

Of course you are correct (not sure why my comment was downvoted) but in the context of having a unique enough hash TAKEN QUICKLY 99.999% of time in set of millions of files its good enough, if one needs better hashing they can hash the whole file but this is quite heavy on large files and pointless if there is no need for it by the application

Hey, this file matches one that the FBI/NSA is looking for via NSL.

Or why not more mundane: another user shares a file with you. Dropbox knows that you already have the same file somewhere on your filesystem outside the Dropbox folder (or a partial match). It doesn't have to transfer that data to you.

But I agree with those saying it's probably a result of some implementation issue (Finder extension or working around some shortcoming in monitoring just the Dropbox directory).

This is pretty close to how their deduplication used to work, and isn't too different than how rsync works.

Of course, this gave rise to the ability to transfer files (even non-public files) quickly between Dropbox accounts provided knowledge of the hashes of its chunks, and Dropbox has since changed their deduplication. See https://github.com/driverdan/dropship

Wouldn't they need to have the file already if they know its hash?

"Hey Dropbox, go find everyone who has a copy of a file with this SHA-256 hash, it's for national security purposes. Thanks"

Speaking of which, does anyone run apps like this with a different user than their own?

I'm thinking of something like

    /home/dropbox drwxrwx--- dropbox <youruser>

Separate from the privacy concerns, what about network traffic? "a few hundred kilobytes" for every file you create adds up.

Any thoughts on SafeMonk, which they claim end to end encryption and piggyback on dropbox?

Uhh... no, no we are not. s/probably //

Dropbox Is Probably Ready To Do So As They Want

I stooped trusting dropbox when condoleeza rice got added to the board.

Surely tradition requires that the title of this piece be "Is Dropbox stealing your files?". I mean, it's a complete waste of a Betteridge event.

Ha, I always wondered if there was a name for this (but obviously also never bothered to Google).

> Betteridge's law of headlines is an adage that states: "Any headline which ends in a question mark can be answered by the word no." It is named after Ian Betteridge, a British technology journalist

It's in response to an earlier-submitted article: https://news.ycombinator.com/item?id=9136546

That is, in fact, titled in the correct manner. Maybe Betteridge doesn't count if the thrust of the article is actually to point out that "no" is the correct answer. I hadn't thought of that.

You upload your files to dropbox servers, how in the world can you come to the delusion that you would notice when they accessed/searched/data mined your files?!

(The article is about uploading files outside your dropbox folder)

Reply on news.ycombinator.com