GCVE-1988-2026-0576
Vulnerability from gna-1988 – Published: 2026-11-19 04:33 – Updated: 2026-11-19 04:33| Vendor | Product | Version | CPE status | |
|---|---|---|---|---|
| Fastly, Inc. | Fastly Next-Gen WAF agent (sigsci-agent) with the NGINX |
Affected:
unknown
|
guessed |
{
"containers": {
"cna": {
"affected": [
{
"product": "Fastly Next-Gen WAF agent (sigsci-agent) with the NGINX",
"vendor": "Fastly, Inc.",
"versions": [
{
"status": "affected",
"version": "unknown"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "finder",
"value": "Zac Lohrenz"
}
],
"descriptions": [
{
"lang": "en",
"value": "Fastly Next-Gen WAF (Signal Sciences) agent: fail-open before rules\nload in Kubernetes sidecar deployments\n\nProduct: Fastly Next-Gen WAF agent (sigsci-agent) with the NGINX\n module, Kubernetes sidecar deployment\nVendor: Fastly, Inc.\nAffected: All agent versions in default configuration, through at\n least 4.83.2 (tested 2026-11-12). Before 4.52.0 no\n mitigation existed. 4.52.0 and later are affected unless\n the startup probe is configured.\nWeakness: CWE-636 Not Failing Securely (\u0027Failing Open\u0027);\n CWE-1188 Initialization of a Resource with an Insecure\n Default; CWE-696 Incorrect Behavior Order\nCVSS 4.0: 6.3 Medium\n CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:L/VA:N/SC:N/SI:L/SA:N\nCVE: Requested from MITRE CNA-LR 2026-09-29; pending\nVendor\nposition: Not a vulnerability; intended fail-open design. Will not\n change the default.\nMitigation: Configure the agent startup probe (4.52.0+)\nResearcher: Zac Lohrenz\nWriteup: https://zaclohrenz.com/posts/cve/fastly-nextgen-waf/\nEvidence: https://github.com/zaclohrenz/fastly-ngwaf-poc\n\n\n1. Summary\n----------\n\nIn the Kubernetes sidecar deployment of Fastly\u0027s Next-Gen WAF (NGINX\ningress controller with the sigsci module, plus a sigsci-agent sidecar\nsharing a Unix socket), the WAF passes traffic uninspected whenever\nthe agent has not loaded a ruleset, while the pod reports Ready.\n\n - Until the agent creates its RPC socket, the module cannot connect\n (errno=2), logs an error, and passes the request through.\n - The agent creates the socket when its first configuration\n download attempt ends, whether or not it succeeded. With no\n ruleset loaded, it answers every RPC.PreRequest with\n WAFResponse 200 and the header \"X-SigSci-Agent: disabled\", and\n keeps doing so until a later download succeeds.\n\nNothing in the default deployment ties pod readiness to rule load.\nEvery rollout, restart, eviction and scale-out creates a window, and\nany startup failure to fetch configuration extends it for as long as\nthe failure lasts.\n\n\n2. Lab testing\n--------------\n\nTested 2026-11-12: sigsci-agent 4.83.2 (index digest\nsha256:4aa2f5296b5f2eeb1a71e79643a93afb19ce85a635a21895ff35b2d0701578d2)\nand sigsci-nginx-ingress-controller 1.15.1-fastly1.4.0 (index digest\nsha256:a4c66bfe5c0b06c6507ce89881ba891711ab3009b2d5d934db192a91e0f452ba),\ndeployed with Fastly\u0027s published sigsci-values.yaml (commit 9794dfb) in\na local kind cluster. Generator at ~20 req/s including SQLi, XSS and\npath traversal payloads; echo backend logging all request headers.\nBlack-box: logs, documented flags, the agent\u0027s documented RPC debug\ndump, and HTTP headers. No Fastly credentials were available, so the\nagent could never load a ruleset.\n\nPhases observed:\n\n Phase Behavior Default signal\n -------------------------- ---------------------- -----------------\n Socket not yet created Module fails open NGINX [error]\n \"cannot connect\n uds socket\" per\n request\n Socket created, no ruleset Agent answers None in NGINX;\n WAFResponse 200 + agent logs only\n X-SigSci-Agent: \"Config update\n disabled failed\" ~every 30s\n\nConfig-download failure types tested, 3 runs each: connection drop,\nHTTP 403 (rejected credentials), DNS NXDOMAIN, TCP reset. All four\nproduced the same state: 20,304 decoded agent responses, one distinct\nvalue (allow, marked disabled). The state persisted with retries every\n~30 s for as long as configuration could not be loaded.\n\nWith slow failures (dropped connections) the socket appeared after\n~60 s and ~935 requests per run passed during the no-socket phase.\nWith fast failures it appeared within ~1.7 s. The 60 s figure is an\nartifact of how egress was blocked, not a typical duration.\n\nMitigations tested:\n\n - Startup probe (SIGSCI_STARTUP_PROBE_LISTENER + startupProbe): pod\n held out of Service endpoints, zero requests to the upstream,\n 3 of 3 runs.\n - Readiness probe initialDelaySeconds 10/30/60 (vendor-suggested\n alternative): traffic, including attack payloads, reached the\n upstream after each delay expired.\n - No documented module or agent option made either component fail\n closed. The experimental debug-standalone=4 removed the\n no-socket phase and the \"disabled\" marker but still allowed every\n request.\n\nFor contrast, freezing the agent after its socket existed (the\ndocumented timeout path) made the module fail open after ~320 ms per\nrequest, also with no log line at default level. The vendor documents\nthe timeout behavior; it documents nothing about an agent that answers\nwith no ruleset loaded.\n\nPer-request evidence of the disabled state appears only with\nnon-default options, e.g. SIGSCI_STATSD_ADDRESS (counter\nsigsci.agent.rpc.tx.dropped.engine_disabled) or\nSIGSCI_DEBUG_LOG_UPLOADS (AgentEnabled:0).\n\nLimits: without working credentials, a healthy start was not observed\n(whether a valid agent ever opens its socket before its first ruleset\nloads, the probe\u0027s 503 -\u003e 200 transition, and how these requests appear\nin the vendor console). The finding is about failure at startup: in\nevery failure tested, an agent that could not load its configuration\nallowed everything until it could, and the default deployment gave\nKubernetes no readiness signal for it.\n\nFull report, manifests, scripts and raw evidence with SHA-256\nchecksums: https://github.com/zaclohrenz/fastly-ngwaf-poc\n\n\n3. Detection\n------------\n\n - Agent container spec: look for SIGSCI_STARTUP_PROBE_LISTENER or\n SIGSCI_STARTUP_PROBE_FILEPATH AND a startupProbe on the agent\n container that checks it. If either is missing, the deployment is\n exposed. The env var alone does nothing.\n - Backend logs: search request headers for \"X-SigSci-Agent:\n disabled\". Requests that passed while the socket was missing carry\n no X-SigSci-* header and will not show up this way.\n - Agent logs: \"Config update failed\" near pod starts.\n - Ingress controller logs: \"cannot connect uds socket\" from the\n sigsci module.\n\n\n4. Mitigation\n-------------\n\nOn agent 4.52.0 or later, configure the startup probe:\n\n containers:\n - name: sigsci-agent\n ports:\n - name: \"startup-port\"\n containerPort: 2024\n env:\n - name: SIGSCI_STARTUP_PROBE_LISTENER\n value: \"0.0.0.0:2024\"\n startupProbe:\n httpGet:\n path: /startup\n port: startup-port\n failureThreshold: 90\n periodSeconds: 2\n\nThe endpoint returns 503 until rules are loaded. Size failureThreshold\nx periodSeconds for the slowest realistic rule fetch; a pod that cannot\nload rules within that time is restarted by the kubelet.\n\nFastly\u0027s published Helm values (sigsci-values.yaml and\nsigsci-nginxinc-values.yaml, commit 9794dfb, 2026-11-10) deploy the\nagent with no startup probe or readiness gate, and no version in that\nrepository\u0027s history has included one. The vendor\u0027s docs-site\nKubernetes guides were updated in October 2026 to include the probe.\n\n\n5. Vendor position\n------------------\n\nFastly\u0027s replies, quoted:\n\n 2026-10-14: Fastly \"does not classify this behavior as a\n vulnerability\"; fail-open is \"a deliberate, foundational design\n choice in the Fastly NGWAF architecture designed to prioritize\n application availability and continuous traffic flow in the event of\n component unavailability, timeouts, or uninitialized states.\" The\n same reply: \"we entirely agree with your assessment of the security\n implications in a dynamically scaling Kubernetes environment.\"\n\n 2026-10-22: \"our install guides and suggested configs don\u0027t include\"\n the startup probe; a CVE is not the right fit \"since the right\n default varies a lot by deployment type (fail-closed at startup is\n ideal for some and an outage for others).\"\n\n 2026-10-29: not a vulnerability under CNA Operational Rules v4.2.6,\n citing 4.1.3.4 (\"CNAs MAY determine consistently well-documented and\n widely known default configurations to not be Vulnerabilities\");\n \"We won\u0027t change the default fail-open behavior in situations where\n the agent is not available yet or is experiencing other issues.\"\n\n 2026-11-03: discussion moved to the CNA-LR; Fastly reserves the right\n to dispute there. Agent versions 4.52.0 and earlier are past its\n two-year support cycle.\n\n 2026-11-17: case closed; Fastly will provide its analysis directly\n to the CNA-LR if the CVE request proceeds.\n\nResearcher position, in brief: CNA rule 4.1.3.2 is a SHOULD for\ndefaults that introduce under-documented or unexpected security\nbehavior, and 4.1.3.3 a SHOULD for defaults that enable direct\nexploitation; 4.1.3.4 is a MAY. The vendor\u0027s reliability documentation\ncovers module fail-open on agent timeout but does not describe an agent\nanswering allow with no ruleset loaded. The vendor stated the agent\n\"still performs preliminary request handling\" while loading its initial\nruleset, but did not answer two written questions (2026-10-30) about\nwhat the agent returns in that state and how operators can detect it.\nThe lab results above answer them as far as possible without\ncredentials.\n\n\n6. Timeline\n-----------\n\n Early 2024 Reported to Fastly support (prior employer\u0027s account)\n 2024-03-14 Agent 4.52.0: startup probe options GA; default\n unchanged\n 2026-04-12 Initial public writeup\n 2026-09-24 Emailed Fastly security requesting classification and\n CVE coordination; no acknowledgment\n 2026-09-29 CVE requested from MITRE CNA-LR\n 2026-10-06 Report forwarded to Fastly support; case opened\n 2026-10-14 Fastly: not a vulnerability; agrees with security\n implications\n 2026-10-22 Fastly: guides omit probe; CVE not the right fit\n 2026-10-29 Fastly final position: not a vulnerability; default\n will not change\n 2026-11-03 Fastly moves discussion to CNA-LR; questions unanswered\n 2026-11-12 Lab testing on agent 4.83.2\n 2026-11-17 Fastly closes case; details published\n Pending CVE assignment\n\n\n7. References\n-------------\n\n Writeup:\n https://zaclohrenz.com/posts/cve/fastly-nextgen-waf/\n Evidence:\n https://github.com/zaclohrenz/fastly-ngwaf-poc\n Agent 4.52.0 changelog:\n https://www.fastly.com/documentation/reference/changes/2024/03/ngwaf-agent-4.52.0/\n Kubernetes startup probe guide:\n https://www.fastly.com/documentation/guides/next-gen-waf/setup-and-configuration/kubernetes/kubernetes-startup-probe/\n Agent reliability documentation:\n \nhttps://www.fastly.com/documentation/guides/next-gen-waf/setup-and-configuration/agent-management/getting-started-with-the-agent/\n sigsci-values.yaml at 9794dfb:\n \nhttps://github.com/signalsciences/sigsci-nginx-ingress-controller/blob/9794dfb38747fb7ec68e4e9286e3164debf1e665/sigsci-values.yaml\n CNA Operational Rules v4.2.6:\n https://www.cve.org/Resources/Roles/Cnas/CNA_Rules_v4.2.6.pdf\n\n--\nZac Lohrenz\n_______________________________________________\nSent through the Full Disclosure mailing list\nhttps://nmap.org/mailman/listinfo/fulldisclosure\nWeb Archives \u0026 RSS: https://seclists.org/fulldisclosure/"
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-1188",
"description": "CWE-1188",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-636",
"description": "CWE-636",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-696",
"description": "CWE-696",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"dateUpdated": "2026-11-19T04:33:41Z",
"orgId": "4e2abfbf-4a2a-4b76-a4e0-d77c18ba156c",
"shortName": "VULNARCHIVE"
},
"references": [
{
"tags": [
"technical-description",
"exploit"
],
"url": "https://vuln.freearchive.org/archive/full-disclosure/2026/Nov/40"
},
{
"tags": [
"technical-description"
],
"url": "https://seclists.org/fulldisclosure/2026/Nov/40"
},
{
"url": "https://github.com/signalsciences/sigsci-nginx-ingress-controller/blob/9794dfb38747fb7ec68e4e9286e3164debf1e665/sigsci-values.yaml"
},
{
"url": "https://github.com/zaclohrenz/fastly-ngwaf-poc"
},
{
"url": "https://nmap.org/mailman/listinfo/fulldisclosure"
},
{
"url": "https://seclists.org/fulldisclosure/"
},
{
"url": "https://www.cve.org/Resources/Roles/Cnas/CNA_Rules_v4.2.6.pdf"
},
{
"url": "https://www.fastly.com/documentation/guides/next-gen-waf/setup-and-configuration/agent-management/getting-started-with-the-agent/"
},
{
"url": "https://www.fastly.com/documentation/guides/next-gen-waf/setup-and-configuration/kubernetes/kubernetes-startup-probe/"
},
{
"url": "https://www.fastly.com/documentation/reference/changes/2024/03/ngwaf-agent-4.52.0/"
},
{
"url": "https://zaclohrenz.com/posts/cve/fastly-nextgen-waf/"
}
],
"source": {
"defect": [
"https://seclists.org/fulldisclosure/2026/Nov/40"
],
"discovery": "EXTERNAL"
},
"title": "Fastly Next-Gen WAF (Signal Sciences) agent fails open before rules load in Kubernetes sidecar deployments (vendor disputes; CVE pending)",
"x_gcve": [
{
"recordType": "advisory",
"relationships": [],
"vulnId": "GCVE-1988-2026-0576",
"x_vulnarchive": {
"archiveUrl": "https://vuln.freearchive.org/archive/full-disclosure/2026/Nov/40",
"automated": true,
"contentSha256": "2241fd939d479606e3c77c1569d11ab47f77032ae558b18e901babbabba14478",
"evidenceScore": 11,
"messageId": "",
"originalUrl": "https://seclists.org/fulldisclosure/2026/Nov/40",
"policy": "vulnarchive-1",
"sourceFormat": "text/html",
"sourcePublishedAt": "2026-11-17T23:29:12Z"
}
}
]
}
},
"cveMetadata": {
"assignerOrgId": "4e2abfbf-4a2a-4b76-a4e0-d77c18ba156c",
"assignerShortName": "VULNARCHIVE",
"datePublished": "2026-11-19T04:33:41Z",
"dateUpdated": "2026-11-19T04:33:41Z",
"state": "PUBLISHED",
"vulnId": "GCVE-1988-2026-0576"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.2"
}
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.