CVE-2026-62239: Improper Link Resolution Before File Access ('Link Following') in Dao-AILab flash-attention
CVE-2026-62239 is a symlink attack vulnerability in Dao-AILab's flash-attention up to version 2.8.3.post1. The issue exists in the download_and_copy() function in hopper/setup.py, which extracts NVIDIA toolchain archives without validating symlinks or filtering tar members. A local attacker can exploit this by pre-planting a symlink in the cache directory to redirect extracted files, enabling arbitrary file writes with the victim's privileges during build time. The vulnerability has a medium severity with a CVSS score of 5.3. No official patch or remediation guidance is currently provided.
AI Analysis
Technical Summary
FlashAttention versions through 2.8.3.post1 contain a vulnerability in the download_and_copy() function within hopper/setup.py. This function extracts NVIDIA toolchain archives but fails to validate symbolic links or filter tar archive members properly. As a result, a local attacker can create a symlink in a predictable cache directory before extraction, causing the extraction process to overwrite arbitrary files chosen by the attacker with the victim's privileges. This improper link resolution before file access (link following) can lead to arbitrary file write during build time. The vulnerability was fixed in a commit identified as 0816ef1, but no patch link or official remediation level is currently documented. The CVSS 4.0 vector indicates local attack vector, low attack complexity, partial privileges required, partial user interaction, and high impact on confidentiality and integrity.
Potential Impact
The vulnerability allows a local attacker to perform arbitrary file writes with the privileges of the victim user during the build process by exploiting symlink handling flaws in archive extraction. This can lead to modification or replacement of critical files, potentially compromising the integrity of the build environment or enabling further attacks. There is no indication of remote exploitation or privilege escalation beyond the victim's privileges. No known exploits are reported in the wild.
Mitigation Recommendations
Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. The vulnerability was fixed in a commit (0816ef1), but no official patch or update release information is provided. Until an official fix is available, avoid running the vulnerable extraction process in untrusted or shared environments, and consider manually verifying or sanitizing archive contents before extraction to prevent symlink attacks.
CVE-2026-62239: Improper Link Resolution Before File Access ('Link Following') in Dao-AILab flash-attention
Description
CVE-2026-62239 is a symlink attack vulnerability in Dao-AILab's flash-attention up to version 2.8.3.post1. The issue exists in the download_and_copy() function in hopper/setup.py, which extracts NVIDIA toolchain archives without validating symlinks or filtering tar members. A local attacker can exploit this by pre-planting a symlink in the cache directory to redirect extracted files, enabling arbitrary file writes with the victim's privileges during build time. The vulnerability has a medium severity with a CVSS score of 5.3. No official patch or remediation guidance is currently provided.
CVSS v4.0
Score 5.3medium
Affected software
Run on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
FlashAttention versions through 2.8.3.post1 contain a vulnerability in the download_and_copy() function within hopper/setup.py. This function extracts NVIDIA toolchain archives but fails to validate symbolic links or filter tar archive members properly. As a result, a local attacker can create a symlink in a predictable cache directory before extraction, causing the extraction process to overwrite arbitrary files chosen by the attacker with the victim's privileges. This improper link resolution before file access (link following) can lead to arbitrary file write during build time. The vulnerability was fixed in a commit identified as 0816ef1, but no patch link or official remediation level is currently documented. The CVSS 4.0 vector indicates local attack vector, low attack complexity, partial privileges required, partial user interaction, and high impact on confidentiality and integrity.
Potential Impact
The vulnerability allows a local attacker to perform arbitrary file writes with the privileges of the victim user during the build process by exploiting symlink handling flaws in archive extraction. This can lead to modification or replacement of critical files, potentially compromising the integrity of the build environment or enabling further attacks. There is no indication of remote exploitation or privilege escalation beyond the victim's privileges. No known exploits are reported in the wild.
Mitigation Recommendations
Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. The vulnerability was fixed in a commit (0816ef1), but no official patch or update release information is provided. Until an official fix is available, avoid running the vulnerable extraction process in untrusted or shared environments, and consider manually verifying or sanitizing archive contents before extraction to prevent symlink attacks.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- VulnCheck
- Date Reserved
- 2026-07-13T16:41:09.007Z
- Cvss Version
- 4.0
- State
- PUBLISHED
- Remediation Level
- null
Threat ID: 6a55561b68715ace43e122cd
Added to database: 07/13/2026, 21:18:19 UTC
Last enriched: 07/13/2026, 21:33:04 UTC
Last updated: 07/13/2026, 21:51:27 UTC
Views: 5
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.