What is XSS?

Cross-site scripting (XSS) is an exploit where the attacker attaches code onto a legitimate website that will execute when the victim loads the website. That malicious code can be inserted in several ways. Most popularly, it is either added to the end of a url or posted directly onto a page that displays user-generated content. In more technical terms, cross-site scripting is a client-side code injection attack. https://www.cloudflare.com/learning/security/threats/cross-site-scripting/

Impact

One-click Lemmy account compromise by social engineering users to click your posts URL.

Reproduction

Lemmy does not properly sanitize URI’s on posts leading to cross-site scripting. You can see this working in action by clicking the “link” attached to this post on the web client.

To recreate, simply create a new post with the URL field set to: javascript:alert(1)//

Patching

Adding filtering to block javascript: and data: URI’s seems like the easiest approach.

  • terribleplan@lemmy.nrd.liEnglish
    12·
    1 year ago

    Damn… seems like there should be filtering to only allow http: and https: URIs…

    Did you try the security email on github? I sent a vulnerability (that actually is way fucking worse than I thought given this issue) over a week ago and have heard nothing, so will be posting publicly soon.

    • terribleplan@lemmy.nrd.liEnglish
      13·
      1 year ago

      Holy shit holy shit holy shit. Serious vulnerability confirmed. Combined with the issue(s) I have tried to report this is insane. I just tested this (and purged it so as not to publicly disclose just yet). This is really bad.

        • Dusty@l.dustybeer.comEnglish
          9·
          1 year ago

          If you find a way to disclose vulnerabilities without being ghosted by Lemmy developers: update me.

          How have you been “ghosted by Lemmy developers” especially if you “do not use GitHub”

          • terribleplan@lemmy.nrd.liEnglish
            6·
            1 year ago

            Yeah, I just wrote this up as a bug on github and added in that I tried to email them and to please get in contact about the other thing. Hopefully they see it. I can understand checking that email being overlooked considering how busy they likely are given the sudden influx and scaling issues.

            • foo@sh.itjust.worksEnglish
              4·
              1 year ago

              It’s been a bit of a busy week for them. Maybe you can cut them some slack and try again?

        • terribleplan@lemmy.nrd.liEnglish
          8·
          1 year ago

          Yeah, I found something that was “holy shit this is bad if someone finds a way to do X” and tried to report that but didn’t dig any deeper. This is X.

    • terribleplan@lemmy.nrd.liEnglish
      2·
      1 year ago

      I believe if unsafe-inline were removed from script-src then the CSP would block this.

      If the frontend depends on inline script tags then this likely can’t be changed super easily… The fact that unsafe-eval is in script-src is kinda worrying as well. Ideally you would lock the CSP down a lot more than they have.

      • hawkwind@lemmy.managementEnglish
        0·
        1 year ago

        Aye, I am pretty sure CSP is bypass-able in most situations unless your pinning checksums or hashes. Just thought it might help take the edge off the hacker panic.

        • terribleplan@lemmy.nrd.liEnglish
          1·
          1 year ago

          Yeah, it can certainly help in some cases, defense in depth and all that. If the CSP were ‘self’ (allowing any JS hosted on your domain) this would probably be DoA. Sadly, until the frontend stops using <script> to set things on window to hydrate state from SSR to client-side they won’t be able to change it without breaking things.