CVE-2026-54917 is a path-traversal vulnerability in SeaweedFS versions before 4.30 affecting the S3 API gateway and Iceberg REST catalog gateway. The gateways construct routers with path cleaning disabled through mux.NewRouter().SkipClean(true). A traversal segment can therefore remain in an object-path variable during routing, while subsequent filer-path construction using util.JoinPath for S3 or path.Join for Iceberg collapses that segment. A request associated with one bucket can consequently be resolved against a different bucket or namespace.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
.., encoded dot segments, and backslash-based traversal. Do not treat disabled authentication as a security boundary.Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
Repository is a self-contained proof-of-concept for CVE-2026-54917 affecting SeaweedFS S3 gateway versions before 4.30. The main exploit is exploit.py, a standalone Python script using only the standard library. It manually constructs HTTP requests and AWS SigV4 signatures, then sends raw GET or PUT requests over a socket so traversal sequences are preserved exactly on the wire. The exploit abuses a path traversal/confused-deputy condition where SeaweedFS authorizes access based on the routed bucket variable but later resolves object paths containing '..' into a different bucket on the backend. Core capability: unauthorized cross-bucket read and write. The script supports four traversal encodings: plain '../', '%2e%2e', '..%2f', and '..%5c'. It can read arbitrary object contents from a target bucket or upload attacker-controlled file contents into that bucket while authenticating only for a different authorized bucket. Repository structure: README.md explains impact, usage, and reproduction; ANALYSIS.md provides root-cause analysis tied to SeaweedFS source behavior; EVIDENCE.txt contains vulnerable vs patched lab results; patch-4.30.diff shows the fix introducing validateRequestPath middleware; lab/docker-compose.yml, lab/s3.json, and lab/setup.sh provide a reproducible environment with SeaweedFS 4.29/4.30, IAM identities, and seeded buckets/objects. This is a real exploit, not merely a detector, and demonstrates both confidentiality and integrity impact over the network against the S3 API endpoint.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
4 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A prior SeaweedFS path-traversal vulnerability whose version 4.30 fix hardened request URL paths but did not protect the X-Amz-Copy-Source header addressed by CVE-2026-55874.
A path traversal vulnerability detection entry related to SeaweedFS, referenced as CVE-2026-54917.
A high-severity path traversal vulnerability in SeaweedFS prior to 4.30. Crafted requests such as GET /bucket-A/../evil-bucket/key can cause the S3 or Iceberg gateway to access objects in an unintended bucket, potentially resulting in cross-bucket confidentiality and integrity impact.
A critical path traversal vulnerability in SeaweedFS that allows unauthenticated remote attackers to bypass bucket isolation and gain arbitrary cross-bucket read, write, or delete access via the S3 and Iceberg REST gateways.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.