GHSA-PFWX-XJP4-QXCJ
Vulnerability from github – Published: 2026-11-18 20:42 – Updated: 2026-11-18 20:42Summary
atomic-server provisions a well-known invite at {server}/setup with write = true, target = the Drive root, and usagesLeft = 1. Invite redemption ignores the redeemer's identity (for_agent: _ in server/src/plugins/invite.rs:22), so possession of the URL is the only requirement. An unauthenticated attacker who sends GET /setup?public-key=<attacker_key> has the server create a new Agent and grant it write access to the entire Drive. On a freshly deployed (not-yet-claimed) instance this is a full server takeover with no credentials.
Severity
Critical — CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H (9.8). No authentication required; grants attacker-controlled write over all resources in the Drive.
Affected component
server/src/appstate.rs— creates the/setupinvite targeting the Drive root withwrite=true,usagesLeft=1.server/src/plugins/invite.rs:14-173—construct_invite_redirectdestructuresfor_agent: _(redeemer identity discarded);?public-key=→Agent::new_from_public_key→add_rights(agent, target, write=true).
Steps to Reproduce
Environment: joepmeneer/atomic-server:develop in Docker, http://localhost:9883, fresh store. No auth headers are used.
-
Confirm the invite is live —
GET /setup(Accept:application/ad+json) showsinvite/write = true,invite/target = http://localhost:9883(Drive root),usagesLeft = 1. -
Redeem it anonymously with an attacker-generated ed25519 public key:
GET /setup?public-key=q2UkizMX0Go4mHwW2BXepqkzpP5FHlagEft8M32t0EM%3D HTTP/1.1
Host: localhost:9883
Accept: application/ad+json
Response 200 OK returns a Redirect confirming the new agent:
{
"https://atomicdata.dev/properties/invite/redirectAgent":
"http://localhost:9883/agents/q2UkizMX0Go4mHwW2BXepqkzpP5FHlagEft8M32t0EM=",
"https://atomicdata.dev/properties/isA": ["https://atomicdata.dev/classes/Redirect"]
}
GET /shows the attacker is now in the Drive'swritearray:
"https://atomicdata.dev/properties/write": [
"http://localhost:9883/agents/+/KwsAmyzc7S/...", // legit root
"http://localhost:9883/agents/q2UkizMX0Go4mHwW2BXepqkzpP5FHlagEft8M32t0EM=" // ATTACKER
]
- The attacker signs a normal Commit (ed25519 over the JCS-canonical commit) and
POST /commit(200) — the Drive root was successfully edited. Write access proven.
Expected: /setup must require operator authentication / a one-time secret.
Actual: Any unauthenticated client on the network redeems it and gains Drive-wide write.
Impact
Unauthenticated write to the entire Drive: read/modify/delete any resource, create/alter Agents, reconfigure permissions. This is a complete takeover on a fresh deployment. (It also serves as the entry point of a larger chain: with write access, an attacker can read the server root private key and impersonate root — reported separately.)
Remediation
- Require a one-time secret (printed to stdout at startup) or operator authentication to redeem
/setup. - Invalidate
/setupafter first use with an atomicusagesLeftdecrement (sled transaction) — the current read-check-decrement is a TOCTOU that allows a single-use invite to be redeemed multiple times concurrently. - Fix the inverted invite-expiry check in
invite.rs(if expires > now { reject }is backwards, so expired invites remain redeemable and valid ones are rejected).
{
"affected": [
{
"package": {
"ecosystem": "crates.io",
"name": "atomic-server"
},
"ranges": [
{
"events": [
{
"introduced": "0.29.0"
},
{
"last_affected": "0.40.1"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-69157"
],
"database_specific": {
"cwe_ids": [
"CWE-306"
],
"github_reviewed": true,
"github_reviewed_at": "2026-11-18T20:42:33Z",
"nvd_published_at": "2026-11-17T21:18:12Z",
"severity": "CRITICAL"
},
"details": "## Summary\n\natomic-server provisions a well-known invite at `{server}/setup` with `write = true`, `target` = the Drive root, and `usagesLeft = 1`. Invite redemption **ignores the redeemer\u0027s identity** (`for_agent: _` in `server/src/plugins/invite.rs:22`), so possession of the URL is the only requirement. An **unauthenticated** attacker who sends `GET /setup?public-key=\u003cattacker_key\u003e` has the server create a new Agent and grant it **write access to the entire Drive**. On a freshly deployed (not-yet-claimed) instance this is a full server takeover with no credentials.\n\n## Severity\n\n**Critical \u2014 CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H (9.8).** No authentication required; grants attacker-controlled write over all resources in the Drive.\n\n## Affected component\n\n- `server/src/appstate.rs` \u2014 creates the `/setup` invite targeting the Drive root with `write=true`, `usagesLeft=1`.\n- `server/src/plugins/invite.rs:14-173` \u2014 `construct_invite_redirect` destructures `for_agent: _` (redeemer identity discarded); `?public-key=` \u2192 `Agent::new_from_public_key` \u2192 `add_rights(agent, target, write=true)`.\n\n## Steps to Reproduce\n\nEnvironment: `joepmeneer/atomic-server:develop` in Docker, `http://localhost:9883`, fresh store. No auth headers are used.\n\n1. Confirm the invite is live \u2014 `GET /setup` (Accept: `application/ad+json`) shows `invite/write = true`, `invite/target = http://localhost:9883` (Drive root), `usagesLeft = 1`.\n\n2. Redeem it anonymously with an attacker-generated ed25519 public key:\n\n```http\nGET /setup?public-key=q2UkizMX0Go4mHwW2BXepqkzpP5FHlagEft8M32t0EM%3D HTTP/1.1\nHost: localhost:9883\nAccept: application/ad+json\n```\n\nResponse `200 OK` returns a `Redirect` confirming the new agent:\n\n```json\n{\n \"https://atomicdata.dev/properties/invite/redirectAgent\":\n \"http://localhost:9883/agents/q2UkizMX0Go4mHwW2BXepqkzpP5FHlagEft8M32t0EM=\",\n \"https://atomicdata.dev/properties/isA\": [\"https://atomicdata.dev/classes/Redirect\"]\n}\n```\n\n3. `GET /` shows the attacker is now in the Drive\u0027s `write` array:\n\n```json\n\"https://atomicdata.dev/properties/write\": [\n \"http://localhost:9883/agents/+/KwsAmyzc7S/...\", // legit root\n \"http://localhost:9883/agents/q2UkizMX0Go4mHwW2BXepqkzpP5FHlagEft8M32t0EM=\" // ATTACKER\n]\n```\n\n4. The attacker signs a normal Commit (ed25519 over the JCS-canonical commit) and `POST /commit` (200) \u2014 the Drive root was successfully edited. Write access proven.\n\n**Expected:** `/setup` must require operator authentication / a one-time secret.\n**Actual:** Any unauthenticated client on the network redeems it and gains Drive-wide write.\n\n## Impact\n\nUnauthenticated write to the entire Drive: read/modify/delete any resource, create/alter Agents, reconfigure permissions. This is a complete takeover on a fresh deployment. (It also serves as the entry point of a larger chain: with write access, an attacker can read the server root private key and impersonate root \u2014 reported separately.)\n\n## Remediation\n\n- Require a one-time secret (printed to stdout at startup) or operator authentication to redeem `/setup`.\n- Invalidate `/setup` after first use with an **atomic** `usagesLeft` decrement (sled transaction) \u2014 the current read-check-decrement is a TOCTOU that allows a single-use invite to be redeemed multiple times concurrently.\n- Fix the inverted invite-expiry check in `invite.rs` (`if expires \u003e now { reject }` is backwards, so expired invites remain redeemable and valid ones are rejected).",
"id": "GHSA-pfwx-xjp4-qxcj",
"modified": "2026-11-18T20:42:33Z",
"published": "2026-11-18T20:42:33Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/ontola/atomic-server/security/advisories/GHSA-pfwx-xjp4-qxcj"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-69157"
},
{
"type": "WEB",
"url": "https://github.com/ontola/atomic-server/commit/98517d8921c78bd9b728a2459378c184ebf8a969"
},
{
"type": "PACKAGE",
"url": "https://github.com/ontola/atomic-server"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "Atomic Server: Unauthenticated `/setup` invite grants write access to the entire Drive"
}
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.