Skip to content

Commit 0607c14

Browse files
authored
Merge pull request #1269 from jasnow/oct-22a-advs
Two new advisories @simi - Thanks for reviewing and approving this PR.
2 parents e838ab4 + a0aa8d1 commit 0607c14

2 files changed

Lines changed: 94 additions & 0 deletions

File tree

‎gems/crass/GHSA-pq4g-8rxf-4g29.yml‎

Lines changed: 42 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,42 @@
1+
---
2+
gem: crass
3+
ghsa: pq4g-8rxf-4g29
4+
url: https://github.com/rgrove/crass/security/advisories/GHSA-pq4g-8rxf-4g29
5+
title: Incorrect tokenization of bad URLs enables CSS sanitizer bypass
6+
date: 2026-10-21
7+
description: |
8+
## Summary
9+
10+
Crass's tokenizer didn't terminate a bad URL token at a ) that followed
11+
an escaped backslash (\\). As a result, Crass consumed more input as
12+
part of the bad URL than a spec-compliant browser does. CSS sanitizers
13+
that parse untrusted CSS with Crass and re-emit it could be tricked
14+
into passing through declarations, style rules, or at-rules (including
15+
@import) that the sanitizer would otherwise have removed, but that
16+
a browser would then parse and apply.
17+
18+
## Impact
19+
20+
Crass by itself only mis-tokenizes input. The security impact is on
21+
CSS sanitizers that use Crass to make allow/deny decisions and then
22+
re-serialize the CSS. Exploitation requires a sanitizer that doesn't
23+
reject bad URL tokens wherever they occur. Confirmed examples:
24+
25+
- Sanitize (all versions through 7.0.0): Sanitize::CSS.properties,
26+
style attributes, Sanitize::CSS.stylesheet, and <style> elements.
27+
- Loofah (Loofah::HTML5::Scrub.scrub_css, used for style attributes),
28+
and therefore rails-html-sanitizer (sanitize_css, and sanitize when
29+
style is a permitted attribute).
30+
cvss_v4: 5.3
31+
patched_versions:
32+
- ">= 1.0.8"
33+
related:
34+
url:
35+
- https://rubygems.org/gems/crass/versions/1.0.8
36+
- https://github.com/rgrove/crass/blob/v1.0.8/HISTORY.md#108-2026-10-21
37+
- https://github.com/rgrove/crass/compare/v1.0.7...v1.0.8
38+
- https://drafts.csswg.org/css-syntax-3/#consume-remnants-of-bad-url
39+
- https://github.com/rgrove/crass/security/advisories/GHSA-pq4g-8rxf-4g29
40+
notes: |
41+
- No CVE in GHSA URL.
42+
- cvss_v4 from GHSA URL.
Lines changed: 52 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,52 @@
1+
---
2+
gem: view_component
3+
ghsa: 8qw7-6phv-7q6p
4+
url: https://github.com/ViewComponent/view_component/security/advisories/GHSA-8qw7-6phv-7q6p
5+
title: ViewComponent - Incomplete fix for CVE-2026-54497 — slot and
6+
with_content state still leak across reused component instances
7+
date: 2026-08-28
8+
description: |
9+
## Summary
10+
11+
CVE-2026-54497 was fixed by having render_in call __vc_reset_render_state!,
12+
which clears exactly three ivars (@__vc_controller, @__vc_helpers,
13+
@__vc_request) on every render so a reused ViewComponent::Base instance
14+
can't leak the previous request's controller/helpers/request. The
15+
fix's own comment admits it deliberately leaves two other pieces of
16+
render-scoped state untouched: @__vc_set_slots (populated by with_<slot>
17+
setters) and @__vc_content_set_by_with_content (populated by with_content),
18+
reasoning that these are "populated by callers before render_in runs."
19+
That reasoning only holds for the intended one-instance-per-render
20+
usage pattern — it breaks down in exactly the reuse/aliasing scenario
21+
the CVE fix was written for, because these two ivars hold the actual
22+
content/data being rendered (often user- or session-specific), not
23+
just framework plumbing.
24+
25+
## IMPACT
26+
27+
Any application that follows the common ViewComponent idiom of
28+
pre-populating a component with request/user data via
29+
with_content/with_<slot> and then ends up reusing that same instance
30+
across two different requests/users — due to caching, pooling,
31+
memoization, or the aliasing pattern the vendor's own fix commit
32+
explicitly anticipated — leaks the first user's data to the second,
33+
despite the CVE-2026-54497 fix having run.
34+
35+
## CREDIT
36+
37+
Yazan Balawneh, Cystack.ps
38+
cvss_v3: 6.8
39+
patched_versions:
40+
- ">= 4.14.0"
41+
related:
42+
url:
43+
- https://github.com/ViewComponent/view_component/blob/main/docs/CHANGELOG.md#4140
44+
- https://github.com/ViewComponent/view_component/compare/v4.13.0...v4.14.0
45+
- https://github.com/ViewComponent/view_component/pull/2691
46+
- https://github.com/ViewComponent/view_component/security/advisories/GHSA-8qw7-6phv-7q6p
47+
notes: |
48+
- No CVE in GHSA URL.
49+
- cvss_v3 from GHSA URL.
50+
- Note that gem name in view_component and repo name is ViewComponent.
51+
- GHSA affected range is "<= 4.12.0", but 4.13.0 is also vulnerable;
52+
the fix shipped in 4.14.0 per the view_component CHANGELOG.

0 commit comments

Comments
 (0)