You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
SixLabors.ImageSharp is vulnerable to Out-of-bounds Write in versions 3.0.0 - 4.1.0 #156
You are affected if you use a version within the vulnerable range and decode untrusted TIFF images that use CCITT/fax (Group 3 or Modified Huffman) compression.
Background info
SixLabors.ImageSharp is vulnerable to an out-of-bounds write in versions 3.0.0-4.1.0.
How to fix this
Upgrade the SixLabors.ImageSharp library to the patched version.
RotateFlip and Redact directly call ImageSharp (AnyBitmap.cs:1486).
CreateMultiFrameTiff(IEnumerable) directly loads input paths through ImageSharp (AnyBitmap.cs:1224).
The normal multi-page TIFF path uses BitMiracle.LibTiff.NET, so that does not reach these ImageSharp bugs. However, the rest of the calls bypass that protection.
The netstandard2.0 build uses ImageSharp 2.1.13 (IronSoftware.Drawing.Common.csproj:43). This is outside the affected range.
To fix this, you need to upgrade the net60 ImageSharp dependency to 4.1.1 or later, and update the matching ImageSharp.Drawing dependency and the manually maintained versions in IronSoftware.Drawing.nuspec.
This issue has already been logged with our development team and is currently under discussion. We're treating it as a high priority, and I'll get back to you as soon as it's resolved.
Adding a related ask, since this ticket is already tied to Jira ticket DW-38 — your internal ticket (referenced in PR #158) for adding a .NET 8 target with the patched ImageSharp 4.1.2.
We're tracking a Snyk finding against SixLabors.ImageSharp on our side (CVE range [2.1.0,4.1.2) / [1.0.0-beta0001,4.1.2) — Out-of-bounds Write, Memory Allocation with Excessive Size, Infinite Loop), which affects the netstandard2.0 build of IronSoftware.System.Drawing pinned to ImageSharp 2.1.13, since we're on .NET Framework and can't move to the .NET 8 target that PR #158 fixes.
@sydseter's comment above confirms the netstandard2.0 build is on 2.1.13 and currently out of scope for this fix. Separately, issue #131 (closed in <24h) shows a precedent where a netstandard2.0-targeting ImageSharp bump (2.1.9 → 2.1.10) was shipped alongside the net60 fix for a different CVE.
Is a similar backport possible here — bumping the netstandard2.0 target to whatever current Apache-licensed SixLabors.ImageSharp 2.1.x release addresses these CVEs (if one exists), separate from the .NET 8/ImageSharp 4.x work in PR #158?
That would let .NET Framework consumers like us close this out without needing a framework migration.
Happy to open a separate issue if that's cleaner for tracking.
Activity
timgilmore commentedon Sep 10, 2026
Aikido also flagged this for us on 8/25/26. We need this to be resolved to satisfy security requirements.
sydseter commentedon Sep 30, 2026
The vulnerable decoder is exploitable as far as I can see.
The netstandard2.0 build uses ImageSharp 2.1.13 (IronSoftware.Drawing.Common.csproj:43). This is outside the affected range.
To fix this, you need to upgrade the net60 ImageSharp dependency to 4.1.1 or later, and update the matching ImageSharp.Drawing dependency and the manually maintained versions in IronSoftware.Drawing.nuspec.
dany-ironsoftware commentedon Oct 8, 2026
Hi @sydseter,
Thanks for the report.
This issue has already been logged with our development team and is currently under discussion. We're treating it as a high priority, and I'll get back to you as soon as it's resolved.
garyrknutson commentedon Oct 17, 2026
Adding a related ask, since this ticket is already tied to Jira ticket DW-38 — your internal ticket (referenced in PR #158) for adding a .NET 8 target with the patched ImageSharp 4.1.2.
We're tracking a Snyk finding against SixLabors.ImageSharp on our side (CVE range [2.1.0,4.1.2) / [1.0.0-beta0001,4.1.2) — Out-of-bounds Write, Memory Allocation with Excessive Size, Infinite Loop), which affects the netstandard2.0 build of IronSoftware.System.Drawing pinned to ImageSharp 2.1.13, since we're on .NET Framework and can't move to the .NET 8 target that PR #158 fixes.
@sydseter's comment above confirms the netstandard2.0 build is on 2.1.13 and currently out of scope for this fix. Separately, issue #131 (closed in <24h) shows a precedent where a netstandard2.0-targeting ImageSharp bump (2.1.9 → 2.1.10) was shipped alongside the net60 fix for a different CVE.
Is a similar backport possible here — bumping the netstandard2.0 target to whatever current Apache-licensed SixLabors.ImageSharp 2.1.x release addresses these CVEs (if one exists), separate from the .NET 8/ImageSharp 4.x work in PR #158?
That would let .NET Framework consumers like us close this out without needing a framework migration.
Happy to open a separate issue if that's cleaner for tracking.