It doesn't work. The person who posted the comment you're responding to has absolutely no idea what he's talking about. He confabulated the entire explanation based on a single misunderstood block of code containing the comment «Remove " - Chrome Web Store" suffix if present» in the (local, NodeJS-powered) scraper that the person who's publishing this data themselves used to fetch extension names.
No. Firefox always randomizes the extension ID used for URLs to web accessible resources on each restart [1]. Apparently, manifest v3 extensions on Chromium can now opt into similar behavior [2].
An additional improvement added in manifest v3 in both Chromium and Firefox is that extensions can choose to expose web accessible resources to only certain websites. Previously, exposing a web accessible resource always made that resource accessible to all websites.
Is there no browser setting to defend against this attack? If not, there should be, versus relying on extension authors to configure or enable such a setting.
I imagine that it would require browsers to treat web requests from JS differently from those initiated by the user, specifically pretending the JS-originating requests are by logged-out or "incognito" users (by, I suppose, simply not forwarding any local credentials along, but maybe there's more to it than that).
Which would probably wreak havoc with a lot of web apps, at least requiring some kind of same-origin policy. And maybe it messes with OAuth or something. But it does seem at least feasible.
I'm not sure how you'd patch that. Any request that’s made from the current open tab / window is made on behalf of the user. From my point of view, it's impossible for the browser to know, if the request is legit or not.
An ideal implementation of the same origin policy would make it impossible for a site (through a fetch call or otherwise) to determine whether an extension resource exists/is installed or the site simply lacks permission to access it.
If this is true, it's insane that this would work:
- why does CWS respond to cross-site requests?
- why is chrome sending the credentials (or equivalent) in these requests?
- why is the button enabled server-side and not via JS? Google must be confident in knowing the exact and latest state of your installed extensions enough to store it on their servers, I guess
It's not true. The person you're responding to has a habit of posting implausible-but-plausibly-plausible nonsense, and it's not how this works at all.
I made the mistake of trying to skim the code hastily before I had to leave to run an errand, and yes it turns out I was wrong, but please refrain from the personal comments, and no, I don't have any such "habit."
Wrong again. (PS: The fact that you have now replied—which automatically disables comment deletion—is the only thing that prevented my removing it just now. So great job.)
> The fact that you have now replied—which automatically disables comment deletion—is the only thing that prevented my removing it just now. So great job.
How was I supposed to know that you intended to delete it?
In any case, you may still have time to edit your comment, as I did with my erroneous root-level comment, since I can't delete that either, for the same reason.
Isn't it enumerating web_accessible_resources? Below static collectFeatures(e, t) there is a mapping of extension IDs to files in the const r (Minified JS, obviously.)
How do you patch it? The extensions themselves (presumably) need to access the same web accessible resources from their content scripts. How do you differentiate between some extension’s content script requesting the resource and LinkedIn requesting it?
The file is then available using a URL like: moz-extension://<extension-UUID>/images/my-image.png"
<extension-UUID> is not your extension's ID. This ID is randomly generated for every browser instance.
This prevents websites from fingerprinting a browser by examining the extensions it has installed.
It does by default, except for the files from the extension that the extension author has explicitly designated as content-accessible. It's explained ("Using web_accessible_resources") at the other end of the link.
Fingerprinting. There are a few reasons you'd do it:
1. Bot prevention. If the bots don't know that you're doing this, you might have a reliable bot detector for a while. The bots will quite possibly have no extensions at all, or even better specific exact combination they always use. Noticing bots means you can block them from scraping your site or spamming your users. If you wanna be very fancy, you could provide fake data or quietly ignore the stuff they create on the site.
2. Spamming/misuse evasion. Imagine an extension called "Send Messages to everybody with a given job role at this company." LinkedIn would prefer not to allow that, probably because they'd want to sell that feature.
I wrote some automation scripts that are not triggered via browser extensions (e.g., open all my sales colleagues’ profiles and like their 4 most recent unliked posts to boost their SSI[1], which is probably the most ‘innocent’ of my use-cases). It has random sleep intervals. I’ve done this for years and never faced a ban hammer.
Wonder if with things like Moltbot taking the scene, a form of “undetectable LinkedIn automation” will start to manifest. At some point they won’t be able to distinguish between a chronically online seller adding 100 people per day with personalized messages, or an AI doing it with the same mannerisms.
Skimming the list, looks like most extensions are for scraping or automating LinkedIn usage. Not surprising as there's money to be made with LinkedIn data. Scraping was a problem when I worked there, the abuse teams built some reasonably sophisticated detection & prevention, and it was a constant battle.
I think they framed it this way because they don't consider scraping abuse (to be fair, neither do I, as long as it doesn't overload the site). Botting accounts for spam is clear abuse, however, so that's fair game.
No, I consider all data collection and scraping egregious. From that perspective, LinkedIn is hypocritical when Microsoft discloses every filesystem search I do locally to bing.
The big social media businesses deserve a Teddy Roosevelt character swooping in and busting their trusts, forcing them to play ball with others even if it destroys their moats. Boo hoo! Good riddance. World's tiniest violin.
This is a popular position across the aisle. Here's hoping the next guy can't be bought, or at least asks for more than a $400M tacky gold ballroom!
We enjoy the fruits of an LLM or two from time to time, derived from hoards of ill gotten data. Linkedin has the resourses to attempt to block scraping, but even at the resource scale of LI I doubt the effort is effective.
I am not denying that scraping is useful. If it wasn't people wouldn't do it. But if the site rules say you aren't allowed to scrape, then I don't think people should be hostile towards the people enforcing the rules.
Well, they can try to enforce the rules; that's perfectly fair. At the same time, there are many methods of "trying" which I would not consider valid or acceptable ones. "Enforcing the rules" does not give a carte blanche right to snoop and do "whatever's necessary." Sony tried that with their CD rootkits and got multiple lawsuits.
Do they respect my data? Why do they get to track me across sites when I clearly don't want them to but someone can't scrape their data when they don't want them to. Why should big companies get the pass but individuals not? They clearly consider internet traffic fair game and are invasive and abusive about it so it is not only fair to be invasive and abusive back, it is self defense at this point.
I'm sure there are issues with fake accounts for scraping, but the core issue is that LinkedIn considers the data valuable. LinkedIn wants to be able to sell the data, or access to it at least, and the scrapers undermine that.
They could stop all the scraping by providing a downloadable data bundle like Wikipedia.
I mean, regardless of who they are or even if you don’t like what LinkedIn does themselves with the data people have given them, the random third parties with the extensions don’t additionally deserve to just grab all that data too, do they?
Eh. I worked at a company which made an extension which scraped LinkedIn. We provided a service to recruiters, who would start a hiring process by putting candidates into our system.
The recruiters all had LinkedIn paid accounts, and could access all of this data on the web. We made a browser extension so they wouldn’t need to do any manual data entry. Recruiters loved the extension because it saved them time.
I think it was a legitimate use. We were making LinkedIn more useful to some of their actual customers (recruiters) by adding a somewhat cursed api integration via a chrome extension. Forcing recruiters to copy and paste did’t help anyone. Our extension only grabbed content on the page the recruiter had open. It was purely read only and scoped by the user.
"The code" here you're referring to (fetch_extension_names.js[1]) isn't and doesn't claim to be LinkedIn's fingerprinting code. It's a scraper that the researcher behind this repo wrote in order to themselves create the CSV of the data that they're publishing.
LinkedIn's fingerprinting code, as the README explains, is found in fingerprint.js[2], which embeds a big JSON literal with the IDs of the extensions it probes for. (Sickeningly enough, this data starts about two-thirds of the way through the file* and isn't the culprit behind the bulk of its 2.15 MB size…)
* On line 34394; the one starting:
const r = [{
id: "aacbpggdjcblgnmgjgpkpddliddineni",
file: "sidebar.html"
In order to create the data source that LinkedIn's extension-fingerprinting relies on to work, someone (at LinkedIn*?) almost certainly violated the Chrome Web Store TOS—by (perversely*) scraping it.
* if LinkedIn didn't get it from an existing data source
> This repository documents every extension LinkedIn checks for and provides tools to identify them.
I get that the CSV lists the extensions, and the tools are provided in order to show work (mapping IDs to actual software). But how was it determined that LinkedIn checks for extensions with these IDs?
Technical writeup from a few weeks ago by a vendor that explains how LinkedIn does it, then boasts that their approach is "quieter, harder to notice, and easier to run at scale":
I didn't find popular extensions like uBlock or other ad blockers.
The list is full of scammy looking data collection and AI tools, though. Some random names from scrolling through the list:
- LinkedGPT: ChatGPT for LinkedIn
- Apollo Scraper - Extract & Export Apollo B2B Leads
- AI Social Media Assistant
- LinkedIn Engagement Assistant
- LinkedIn Lead Magnet
- LinkedIn Extraction Tool - OutreachSheet
- Highperformr AI - Phone Number and Email Finder
- AI Agent For Jobs
These look like the kind of tools scummy recruiters and sales people use to identify targets for mass spamming. I see several AI auto-application tools in there too.
LinkedIn itself provides tools for scummy recruiters to mass spam, so this is just them protecting their business.
Also, not all of them are data collection tools. There are ad blockers listed (Hide LinkedIn Ads, SBlock - Super Ad Blocker) and just general extensions (Ground News - Bias Checker, Jigit Studio - Screen Recorder, RealEyes.ai — Detect Deepfakes Across Online Platforms, Airtable Clipper).
On Firefox, web accessible resources are available at "moz-extension://<extension-UUID>/myfile.png" <extension-UUID> is not your extension's ID. This ID is randomly generated for every browser instance. This prevents websites from fingerprinting a browser by examining the extensions it has installed. https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/Web...
Doesn't the idea of swapping extension specific IDs to your browser specific extension IDs mean that instead of your browser being identifiable, you become identifiable?
I mean, it goes from "Oh they have X, Y , and Z installed" to "Oh, it's jim bob, only he has that unique set of IDs for extensions"
Another thing... they alter the localStorage & sessionStorage prototype, by wrapping the native ones with a wrapper that prevent keys that not in their whitelist from being set.
I’m probably on the list. I made a LinkedIn Redactor that allowed you to add keywords and remove posts from your thread that included such words. It’s the X feature but for LinkedIn. Anyway, got a cease and desist from those lame fucks at LI. So I removed from the chrome store but it’s still available on GitHub.
Chrome is the new IE6. Google set themselves up to be the next Microsoft and is "ad friendly" in all the creepy ways because that's what Google IS an ad company. All they've contributed to security is diminishing the capability of adblockers and letting malware to do bad things to you as consumers.
lapcat | 2 hours ago
ronsor | 2 hours ago
(Alternatively extension developers can modify their extensions to block these requests!)
0cf8612b2e1e | 2 hours ago
Does Firefox have a similar weakness?
burkaman | an hour ago
Edit: Can't find much documentation on exactly how the anti-fingerprinting works, but this page implies that the browser blocks extension detection: https://support.mozilla.org/en-US/kb/trackers-and-scripts-fi...
cxr | an hour ago
tech234a | an hour ago
[1]: https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/Web...
[2]: https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/Web...
tech234a | 17 minutes ago
toomuchtodo | 2 hours ago
zahlman | 2 hours ago
Which would probably wreak havoc with a lot of web apps, at least requiring some kind of same-origin policy. And maybe it messes with OAuth or something. But it does seem at least feasible.
circuit10 | an hour ago
Browsers already have strong protections against that sort of thing, look up the same-origin policy and CORS
zahlman | an hour ago
MrGilbert | 2 hours ago
ronsor | 2 hours ago
cobertos | 2 hours ago
halapro | 2 hours ago
- why does CWS respond to cross-site requests?
- why is chrome sending the credentials (or equivalent) in these requests?
- why is the button enabled server-side and not via JS? Google must be confident in knowing the exact and latest state of your installed extensions enough to store it on their servers, I guess
cxr | an hour ago
lapcat | 57 minutes ago
cxr | 45 minutes ago
lapcat | 32 minutes ago
How was I supposed to know that you intended to delete it?
In any case, you may still have time to edit your comment, as I did with my erroneous root-level comment, since I can't delete that either, for the same reason.
cxr | 17 minutes ago
usefulposter | 2 hours ago
Edit: Confirmed. It's not pinging the Chrome Web Store. https://blog.castle.io/detecting-browser-extensions-for-bot-...
jsheard | 2 hours ago
Should be patched nonetheless though, that's a pretty obscene fingerprinting vector.
what | an hour ago
jsheard | an hour ago
zahlman | an hour ago
Why can't it just deny access to the specified path, except to the extension itself?
cxr | an hour ago
chocolatkey | 2 hours ago
minkeymaniac | 2 hours ago
Screenshots found here https://x.com/DenisGobo/status/2018334684879438150
9021007 | 2 hours ago
mongrelion | 2 hours ago
jppope | 2 hours ago
Nextgrid | 2 hours ago
HPsquared | 2 hours ago
staticshock | 2 hours ago
Aurornis | an hour ago
I would guess this is for rate limiting and abuse detection.
CobrastanJorji | an hour ago
1. Bot prevention. If the bots don't know that you're doing this, you might have a reliable bot detector for a while. The bots will quite possibly have no extensions at all, or even better specific exact combination they always use. Noticing bots means you can block them from scraping your site or spamming your users. If you wanna be very fancy, you could provide fake data or quietly ignore the stuff they create on the site.
2. Spamming/misuse evasion. Imagine an extension called "Send Messages to everybody with a given job role at this company." LinkedIn would prefer not to allow that, probably because they'd want to sell that feature.
3. User tracking.
xz18r | 11 minutes ago
Wonder if with things like Moltbot taking the scene, a form of “undetectable LinkedIn automation” will start to manifest. At some point they won’t be able to distinguish between a chronically online seller adding 100 people per day with personalized messages, or an AI doing it with the same mannerisms.
[1] https://business.linkedin.com/sales-solutions/social-selling...
rdoherty | 2 hours ago
hsbauauvhabzb | 2 hours ago
charcircuit | 2 hours ago
ronsor | 2 hours ago
hsbauauvhabzb | an hour ago
dylan604 | 30 minutes ago
schmidtleonard | 2 hours ago
This is a popular position across the aisle. Here's hoping the next guy can't be bought, or at least asks for more than a $400M tacky gold ballroom!
sellmesoap | 2 hours ago
charcircuit | an hour ago
ronsor | an hour ago
jmward01 | an hour ago
hsbauauvhabzb | an hour ago
missingdays | an hour ago
zelphirkalt | an hour ago
nitwit005 | an hour ago
They could stop all the scraping by providing a downloadable data bundle like Wikipedia.
compiler-guy | an hour ago
The data bundle doesn't help that at all.
b112 | an hour ago
xp84 | 2 hours ago
mathfailure | an hour ago
ronsor | an hour ago
bigfishrunning | an hour ago
josephg | an hour ago
The recruiters all had LinkedIn paid accounts, and could access all of this data on the web. We made a browser extension so they wouldn’t need to do any manual data entry. Recruiters loved the extension because it saved them time.
I think it was a legitimate use. We were making LinkedIn more useful to some of their actual customers (recruiters) by adding a somewhat cursed api integration via a chrome extension. Forcing recruiters to copy and paste did’t help anyone. Our extension only grabbed content on the page the recruiter had open. It was purely read only and scoped by the user.
hsbauauvhabzb | 51 minutes ago
bryanrasmussen | 2 hours ago
cxr | an hour ago
LinkedIn's fingerprinting code, as the README explains, is found in fingerprint.js[2], which embeds a big JSON literal with the IDs of the extensions it probes for. (Sickeningly enough, this data starts about two-thirds of the way through the file* and isn't the culprit behind the bulk of its 2.15 MB size…)
* On line 34394; the one starting:
1. <https://github.com/mdp/linkedin-extension-fingerprinting/blo...>2. <https://github.com/mdp/linkedin-extension-fingerprinting/blo...>
cxr | an hour ago
* if LinkedIn didn't get it from an existing data source
winddude | an hour ago
zahlman | 2 hours ago
I get that the CSV lists the extensions, and the tools are provided in order to show work (mapping IDs to actual software). But how was it determined that LinkedIn checks for extensions with these IDs?
And is this relevant for non-Chrome users?
usefulposter | 2 hours ago
https://blog.castle.io/detecting-browser-extensions-for-bot-...
Aurornis | an hour ago
I didn't find popular extensions like uBlock or other ad blockers.
The list is full of scammy looking data collection and AI tools, though. Some random names from scrolling through the list:
- LinkedGPT: ChatGPT for LinkedIn
- Apollo Scraper - Extract & Export Apollo B2B Leads
- AI Social Media Assistant
- LinkedIn Engagement Assistant
- LinkedIn Lead Magnet
- LinkedIn Extraction Tool - OutreachSheet
- Highperformr AI - Phone Number and Email Finder
- AI Agent For Jobs
These look like the kind of tools scummy recruiters and sales people use to identify targets for mass spamming. I see several AI auto-application tools in there too.
NicuCalcea | 35 minutes ago
Also, not all of them are data collection tools. There are ad blockers listed (Hide LinkedIn Ads, SBlock - Super Ad Blocker) and just general extensions (Ground News - Bias Checker, Jigit Studio - Screen Recorder, RealEyes.ai — Detect Deepfakes Across Online Platforms, Airtable Clipper).
cbsks | an hour ago
This works by looking for web accessible resources that are provided by the extensions. For Chrome, these are are available in a webpage via the URL chrome-extension://[PACKAGE ID]/[PATH] https://developer.chrome.com/docs/extensions/reference/manif...
On Firefox, web accessible resources are available at "moz-extension://<extension-UUID>/myfile.png" <extension-UUID> is not your extension's ID. This ID is randomly generated for every browser instance. This prevents websites from fingerprinting a browser by examining the extensions it has installed. https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/Web...
awesome_dude | an hour ago
Doesn't the idea of swapping extension specific IDs to your browser specific extension IDs mean that instead of your browser being identifiable, you become identifiable?
I mean, it goes from "Oh they have X, Y , and Z installed" to "Oh, it's jim bob, only he has that unique set of IDs for extensions"
triceratops | an hour ago
awesome_dude | an hour ago
edit: er, I think that that also suggests that I need to restart firefox more often...
tech234a | 25 minutes ago
b112 | an hour ago
I presume the extension knows when it wants to access resources of its own. But random javascript, doesn't.
maples37 | an hour ago
b112 | 48 minutes ago
It won't disclose how, as it says it has had several users report it. And that it expects 50% of the bounty, and will use it for GPU upgrades.
rchaud | an hour ago
dana321 | an hour ago
hasperdi | an hour ago
You can try this by opening devtools and setting
tech234a | an hour ago
xnx | 38 minutes ago
shouldnt_be | an hour ago
https://javascript.plainenglish.io/the-extensions-you-use-ar...
jmholla | an hour ago
DrStartup | an hour ago
Typical early hooks: • fetch wrapper • XMLHttpRequest.prototype.open/send wrapper • WebSocket constructor wrapper • history.pushState/replaceState wrapper • EventTarget.addEventListener wrapper (optional, heavy) • MutationObserver for DOM diffs • Error + unhandledrejection capture
shj2105 | an hour ago
HumanOstrich | 35 minutes ago
input_sh | an hour ago
unstatusthequo | an hour ago
avastel | an hour ago
https://blog.castle.io/detecting-browser-extensions-for-bot-...
bastard_op | 52 minutes ago
0xbadcafebee | 49 minutes ago
themafia | 49 minutes ago
Google became a monopoly. All monopolies do this.
dwedge | 40 minutes ago
ta988 | 37 minutes ago
mrkramer | 32 minutes ago
DOM100 | 18 minutes ago
const msg = createDoneMessage(); msg.style.opacity = '1';