Chrome Private Mode Not Quite Private
wiplash writes "Google Chrome appears to store at least some information related to, and including, the sites that you have visited when browsing in Incognito mode. Lewis Thompson outlines a set of steps you can follow to confirm whether you are affected. He has apparently reported this to Google, but no response has yet been received."
Google is addicted to your information, and will do whatever they can to get more.
They cannot help themselves.
Resist.
If only we could observe something, without effecting it. Oh well.....
Waiting for the other shoe to...
My girlfriend is using Facebook in Incognito mode...
:: There is no light at the end of a tunnel. There is a tunnel after a tunnel : Thom Y.
You mean someone knows when I put my browser in Porn Mode?
using 4.1.249.1064 on Win7.
Try running a strings against places.sqlite in Firefox as well after all the personal history has been cleared - I sometimes see URLs left in there.
One assumes that Mrs. Coward has seen enough of your posts over the years. I'd be surprised if she
even uses the internet anymore.
Creationist Textbook Stickers Declared Unconstitutional by CowboyNeal
all incognito windows share the same session
So, since the example in TFA didn't restart Chrome between incognito windows, I decided to see what happened when I followed the steps with "4.5 Exit chrome completely, then restart", and can confirm that even when Chrome fully exits and is restarted, it remembers the zoom level used in a URL only ever visited in an incognito window.
"The urge to save humanity is almost always a false front for the urge to rule." --H.L. Mencken
Maybe it's an honest mistake. Maybe. We'll find out with how Google reacts to this discovery.
And like many of the comments in TFA, it didn't work for me (using 4.1.249.1064) once I completely closed out chrome.
It seems that the issue only affects certain versions of Chrome... I'm guessing this is an honest bug, but since it's google, everyone freaks the hell out.
A Google App that collects information, even if you ask it not to? Say it isn't so!
"No one is more miserable than the person who wills everything and can do nothing." -Emperor Claudius 10 BC - AD 54
Exactly as reported.
I'm using 5.0.375.29 beta on an Air running 10.6.3 over wifi.
Went to cheese.com (the #1 resource for cheese!) and the zoom held.
Additionally, when I opened a new tab in non-incognito mode, the zoom STILL held, so there is definitely some communication between regular and incognito windows.
I'm devastated that my secret cheese browsing is now public.
"The pie shall be cut in half and each man shall receive.....death. I'll eat the pie."
This isn't even an issue of trust. It's not a question of whether Google is stealing information about you, or even privacy. It's an error or a possible bug wherein the mode where the browser is in essentially *no history* mode isn't working 100% w/o history.
If this is true, then it raises issues of quality control, not trust
...I'm sure enough people already know exactly what information of your doings the browser sends back to Google.
Honest mistake just like the WIFI data collection ordeal? yeah sure ...
There's no mistakes like that happening with Google, only closing data collection after public outrage and blaming it as a mistake
Pulsed Media Seedboxes
There's always Chromium; I run it on Ubuntu. For Windows there's SRWare Iron. I'm not sure which is the preferred build for OSX; perhaps Crossover Chromium. TFA doesn't say whether Chromium is affected. Some comments under TFA state that the effect lasts only until Chrome is restarted, suggesting that the information is stored only in the memory cache.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
This and many other things about privacy concern me. I work at MIT and google and other big companies hang around, and both within academia and industry there are not enough people advocating privacy and information ownership. Trust me, or not, but Big companies lust over personal information.
There's always Firefox, too.
Here's the bug in question, filed about 2 weeks ago:
http://code.google.com/p/chromium/issues/detail?id=43107
Seems like someone looked at it, prioritized and classified it (eg pri-2, internals-cookies).
What's the big deal? It's just a bug that needs to get fixed, not a huge conspiracy by Google.
The remember zoom was added to the 5.x Beta / Dev channels some time ago, and isn't a part of the current Chrome stable build. [ Google Blog Link : http://googlesystem.blogspot.com/2010/05/10-things-to-try-in-google-chrome-5.html ] Nevertheless, I doubt this is sending any information to Google. You forget Chromium is open source.
There are many ways to finger print something that are not reversible. For instance, this is just page viewing preference data about a site you visited. What if it takes a hash of the url and uses that to store settings like current zoom and scroll location. There is almost no way this violates the idea of 'incognito' mode.
How is this addiction any different from, let's say, the phone company?
There's no -1 for "I don't get it."
If someone's dick ends up in your ass, would consider the possibility that it was an honest mistake?
You can't handle the truth.
TFA only mentions zoom levels as being stored -- not any other info from users' porn-mode browsing session, just zoom levels. Chrome recently began saving users' zoom levels (if I'm not mistaken) so that pretty much explains that (while conveniently also accounting for why users of earlier versions may not experiencing this phenomenon as well.) We're all waiting for google to slip up monumentally (or "pull a facebook," if you will,) but unfortunately we'll have to wait another day.
Google is a marketing/sales/advertising company. They can only be trusted to a certain point. Their motives are not those of a generous and altruistic organization. Their motives are consistent with those of the type of business they are. It is as simple as that.
Incompetence. Malice. Sufficiently advanced. Blah blah blah.
Colorless green Cthulhu waits dreaming furiously.
There are also posts that it *does* work on later versions.
Hopefully you will now get modded into oblivion showing that the modding system actually works, so I can truly say:
Well done Slashdot!
I don't get the flap over the wifi collection thing. It was publicly open wifi stuff they were collecting. If I stick a bullhorn out my window and I yell, I'm eating breakfast now, I'm showering now, I'm going to work, is it reasonable to reserve the right to be offended when people know about the particulars of my day?
but have you considered the following argument: shut up.
Troll rated? Really?
Be aware of the version you're using. Chrome v4 *may* not save the zoom level, so it wouldn't show it anyway. I'm on the dev channel, and thus am using the newly-released v6, and it's definitely reproducible.
Submitted by rcamans on Friday October 23 2009, @01:21PM
rcamans writes "Visit a bunch of sites in Chrome incognito, and then look at your history in IE 7. Oh My God! A few of the sites you did not want in history are in IE history? How did they get there? A nasty in Windows XP OS. Oh, man...
These sites do not show in Opera history, Safari history, Chrome history, or FIrefox history. So maybe it has to do with IE integration into the Windows OS. Do not trust Chrome incognito until this bug is fixed. If it can be fixed.
Also, IE7 search history shows Chrome incognito search items. Oops
wake up and hold your nose
The SRWare Iron link is dead.
There is a war going on for your mind.
I have the Chrome 5.0.375.38 beta from Ubuntu 10.04. Browsing Incognito appears to still change a number of files on disk, though I haven't investigated what is changed or stored. Finding the zoom problem is straightforward, though:
Per-site zoom levels are stored in a Preferences file (.config/google-chrome/Default/Preferences for me) in a "per_host_zoom_levels" section. It appears that the key is the domain name and the value is the zoom level. These seem to be saved when Chrome exits and, at least in my version, are set and accessed from both regular and Incognito mode.
So, anyone who can read this file knows on what domains you have set non-default zoom levels, regardless of whether you accessed the site in Incognito mode.
In linux, the zoom preferences are stored in the file ~/.config/chromium/Default/Preferences . Making the default directory non-writable by the user will prevent the zoom level (and whatever else) from being stored.
Chrome is very likely to hold the DOM of visited pages in the cache so that f.e. hitting the back button will quickly render the previous page. That does not necessarily mean that the information gets persisted on the hard drive or is available to other pages. On the other hand it's not unlikely that the information sometimes gets paged out to the hard drive and persists until it gets overwritten.
I just tested it and it works here... in Chromium on Ubuntu Lucid x64, FWIW (not much)
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Terrible example. Collecting wifi ssids doesn't require connecting to the wifi point at all.
It's like recording people's door numbers from a distant except you only know that number is around that area.
Also it IS useful as it allows you to do geolocation in areas where you can't get gps or you want a more accurate gps coord.
Did you even look in options? Turn off "search suggestions". That's the feature that relies on this information being sent to Google.
Please, please stop spreading Microsoft's FUD.
Don't thank God, thank a doctor!
Yeah, it's back up now, was giving a PHP error.
There is a war going on for your mind.
It's because chrome retrieves a list of popular web addresses matching your search. The same thing happens on the google main page with auto complete.
I've noticed that previously visited sites still flash up as suggestions immediately after purging the history. These seems to go away after a page refresh. There's probably some caching going on that isn't deleted correctly.
Fleur de Sel
Come on, people -- we even take a sane position towards Microsoft these days.
Chromium is an open-source project. Write a patch and see what happens.
And if you really insist it must be deliberate, please explain how spying on your fucking zoom level, and storing it in a local file which is never sent over any network, is so dangerous.
Don't thank God, thank a doctor!
If someone's dick ended up in my rectum, I also wouldn't consider the possibility that it's in any way analogous to someone maybe spying on my fucking zoom levels.
Don't thank God, thank a doctor!
Same on my Windows machine. Looks like an oversight in a new feature. That's the risk of using the beta channel, I guess.
Iron works on Linux as well, not just Windows. I run it on Ubuntu 9.10. As I mentioned above, 4.0.275.2 (Developer Build 35171) of Iron is affected by the bug from the article.
Run Firefox or Google Chrome for a few days, click "Clear Recent History", select "Forever", exit them.
Now go to a directory where they store profile data and discover SQLite files containing information from all the web sites you've visited (`man strings`).
Both browsers 'forget' to run VACUUM on SQLite databases they are using. However it would be even better to zero fill all the files containing your traces, then delete 'em, then recreate them.
Oh really? I think it makes perfect sense for someone to spy on your zoom levels to adjust the timing of the next event accordingly.
You can't handle the truth.
I think you're missing the GPs point. Although many around here might well hold the beliefs you allude to (I don't think its a significant population on Slashdot, as victimized as you might feel by them), the GPs point is that the cost of betrayal by the Government far exceeds the cost of betrayal by a Corporation. In fact, the worst a Corporation can do do you is really limited by what the Government will allow it to do - if you are really so afraid of what a Corporation can do to you, you are implicitly afraid of what the Government will let it do.
...when I completely exit chrome and re-visit the same site.
Unless iron is using beta code these days you'll need chromium for proper javascript blocking support. That should be in mainline soon enough, though. Or hell, it might be already, my news is a couple days old now :/
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Using Windows 7 32bits, Chrome version 6.0.401.1
"A sysadmin is a cross between a detective, a police officer, a gardener, a doctor and a fireman"
Not only that, but just because it remembers settings does not prove it remembers the actual domain name. (sha1 of URL name would take care of that...)
That would have been a way better blog post (from a technical aspect). That being said, there are easy ways to fix that.
I'm running chromium on OS X. There is native os x daily build. This is what I use.
Also notice that incognito mode leaves a trail of other content on the system and does not delete it after you close the browser session. For instance say you are looking for videos on "how to prepare meat" and find this fine cooking site. You decide to watch the sample video and it is the perfect meal choice for your surprise for that "significant other". Later on "significant other" happens to look in the /tmp directory. There are video files there and when played reveal the secret meal you are going to prepare. In fact all of the temporary content from your incognito session is neatly stored in one place.
There's always Chromium; I run it on Ubuntu. For Windows there's SRWare Iron. I'm not sure which is the preferred build for OSX; perhaps Crossover Chromium. TFA doesn't say whether Chromium is affected. Some comments under TFA state that the effect lasts only until Chrome is restarted, suggesting that the information is stored only in the memory cache.
Chromium is the exact same code as Chrome. Of course it will be affected, unless you patch it out.
And seriously, people, get a grip. It's not like this tells Google what sites you visited. It just stores it on your computer. It has nothing to do with Google wanting your information, that's a complete non sequitur here. Geez.
MediaWiki developer, Total War Center sysadmin
find -type f -exec grep -i {} /dev/null \;
This also shows the file name.
Iron 5.0.377 for Linux: http://www.srware.net/forum/viewtopic.php?f=18&t=1502
No it would not take care of it since it is relatively easy to generate SHA-1 hashes of URL's. You can then just compare with the one used for the settings. For a wife it would be pretty simple to prove that somebody went to the www.verynastypron.com by simply generating the correct hash. It's obfuscation at best. Of course, it's still not so bad as banks claiming they don't know the PIN of a bank card because they only stored the hash, but it's easy to brute force.
But that's what you use Twitter for!
Just reproduced it with Chrome 6.0.401.1 dev on windows. It remembers your zoom level even if you close chrome completely, making sure there are no chrome.exe processes running, then start chrome back up. Just because it doesn't affect some older versions does not mean this story is false.
sig? uhh, umm, ok
half of the people posting here probably didn't read the article and are going off about google when all that is stored is the zoom level. how can anyone genuinely be concerned about this?
Chromium is the exact same code as Chrome. Of course it will be affected, unless you patch it out.
Chrome is Chromium plus even more monitoring code. Er, I mean, integration, of course. Heh heh.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
http://code.google.com/p/chromium/issues/detail?id=43107
Let's credit them with salting the URL...
Everyone mentioning SRWare Iron should know about this little tidbit: The story of Iron. The article and the linked IRC log tell a very interesting story about a guy less concerned with having a good reason to fork and more concerned with making money off of adsense and publicity for creating a "privacy-respecting" Chrome which is basically a perpetually outdated Chromium with a few checkboxes in "Under the Hood" defaulting to off.
The guy who runs that blog does not try to hide the fact that he's a Chrome developer, and he admits that there is the highly unlikely possibility that the person who was asking these questions was not the person who went on to release Iron. I was skeptical as well until I checked out the log file itself and quite honestly it would have to be an incredible coincidence for this guy to be asking such questions and providing the information that he does in his attempts to glean information on the right way to advertise his product as well as how to go about renaming the executable. There's more that makes it very reasonable to believe this is the guy who went on to release Iron, so please don't dismiss it until you've checked out the log file in detail. If this was a supremely unnecessary and elaborate hoax it sure is pulled off convincingly.
Using Iron after reading this information made me feel like I was supporting the wrong guy here and I couldn't do it anymore, it was just too uncomfortable seeing that this guy was looking for adsense revenue and to make a name for himself. The attitude of this developer is not one I would encourage at all.
"We invented personal computing." - Bill Gates
I use Iron, but not as a general browser. I "firewall" my email and bank web access to the Iron browser and NEVER use it to access any other site. Converesly use Firefox for everything else and NEVER access my email or bank sites with it.
That does not help against brute force attacks that just calculate the hashes, it only helps against attacks that use rainbow tables. The salt must be available, so you can still calculate the hash result.
Anyway, you can still see the difference when you simply visit the page, as others have pointed out.
You could instead (or additionally) use a portable "install" of Iron for each site, preventing any bugs which might leak information to other sites.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Anyway, you can still see the difference when you simply visit the page, as others have pointed out.
True, though less useful if you've infiltrated or seized a hard drive and are trying to find history through read-only operations. Also not useful if someone has only visited a site but not customized it -- assuming it still gets an entry then.
That does not help against brute force attacks that just calculate the hashes, it only helps against attacks that use rainbow tables. The salt must be available, so you can still calculate the hash result.
I suppose the determined hacker could extract the salt from the chrome executable; which would allow brute force as an option. However, it would at least foil the casual family snoop - I suppose that's misleading, as you would think that "incognito mode" is 100% incognito.