GHSA-74VM-87HJ-R66F
Vulnerability from github – Published: 2026-10-16 19:58 – Updated: 2026-10-16 19:58Summary
ValidateResponse panics on a nil pointer when a spec declares a response header with content instead of schema. doc.Validate() accepts that document, so the crash comes from a spec the library just called valid.
Same class as GHSA-jpcw-4wr7-c3vq, which covered the request-parameter case in v0.144.0; this is the response-header path.
Affects github.com/getkin/kin-openapi <= v0.145.0. Reproduced on master 947498e5a6bd, Go 1.26.
Details
A content-typed response header has headerRef.Value.Schema == nil.
validate_response.go:181 passes it straight through:
if decodedValue, found, err = decodeValue(dec, headerName, sm, headerRef.Value.Schema, headerRef.Value.Required); err != nil {
and req_resp_decoder.go:273 is where it lands:
if len(schema.Value.AllOf) > 0 {
Two things I checked rather than assumed. The response does not need to carry the header; declaring it is enough, and then every call for that operation panics. And the body path is fine, :114 returns early on a nil schema, so this is headers only. The PoC keeps the body case as a control.
PoC
Drop into openapi3filter/ and run:
go test ./openapi3filter/ -run NilSchema -v
package openapi3filter_test
import (
"context"
"io"
"net/http"
"net/http/httptest"
"strings"
"testing"
"github.com/getkin/kin-openapi/openapi3"
"github.com/getkin/kin-openapi/openapi3filter"
"github.com/getkin/kin-openapi/routers"
)
const specBody = `
openapi: 3.0.3
info: {title: poc, version: "1.0.0"}
paths:
/x:
get:
responses:
"200":
description: ok
content:
application/json: {}
`
const specHeader = `
openapi: 3.0.3
info: {title: poc, version: "1.0.0"}
paths:
/x:
get:
responses:
"200":
description: ok
headers:
X-Thing:
content:
application/json: {}
`
func validatedInput(t *testing.T, spec string, hdr http.Header) *openapi3filter.ResponseValidationInput {
t.Helper()
doc, err := openapi3.NewLoader().LoadFromData([]byte(spec))
if err != nil {
t.Fatalf("load: %v", err)
}
if err := doc.Validate(context.Background()); err != nil {
t.Fatalf("doc.Validate rejected the spec, so this repro proves nothing: %v", err)
}
op := doc.Paths.Find("/x").Get
req := httptest.NewRequest(http.MethodGet, "/x", nil)
return &openapi3filter.ResponseValidationInput{
RequestValidationInput: &openapi3filter.RequestValidationInput{
Request: req,
Route: &routers.Route{Spec: doc, Operation: op, Method: http.MethodGet, Path: "/x"},
},
Status: 200,
Header: hdr,
Body: io.NopCloser(strings.NewReader(`{}`)),
}
}
// Control: body path returns early on a nil schema.
func TestControl_ResponseBodyNilSchema(t *testing.T) {
in := validatedInput(t, specBody, http.Header{"Content-Type": {"application/json"}})
defer func() {
if r := recover(); r != nil {
t.Errorf("body path panicked, expected clean return: %v", r)
}
}()
if err := openapi3filter.ValidateResponse(context.Background(), in); err != nil {
t.Errorf("body path returned %v, expected nil", err)
}
}
// The response does not carry X-Thing. Declaring the header is enough.
func TestResponseHeaderNilSchema(t *testing.T) {
in := validatedInput(t, specHeader, http.Header{"Content-Type": {"application/json"}})
if err := openapi3filter.ValidateResponse(context.Background(), in); err != nil {
t.Fatalf("no panic; got err = %v (already fixed?)", err)
}
}
Output:
=== RUN TestControl_ResponseBodyNilSchema
--- PASS: TestControl_ResponseBodyNilSchema (0.00s)
=== RUN TestResponseHeaderNilSchema
--- FAIL: TestResponseHeaderNilSchema (0.00s)
panic: runtime error: invalid memory address or nil pointer dereference [recovered, repanicked]
[signal SIGSEGV: segmentation violation code=0x1 addr=0x20 pc=0x7f6d2a]
github.com/getkin/kin-openapi/openapi3filter.decodeValue(...)
/src/openapi3filter/req_resp_decoder.go:273 +0x4a
github.com/getkin/kin-openapi/openapi3filter.validateResponseHeader(...)
/src/openapi3filter/validate_response.go:181 +0x146
github.com/getkin/kin-openapi/openapi3filter.ValidateResponse(...)
/src/openapi3filter/validate_response.go:89 +0xafa
The FAIL is the point: control passes, the header case takes the binary down.
Ran in a container against a fresh clone of master. I have not tried it against a live proxy.
Impact
A process crash instead of an error return. Declaring the header is enough, so it is every call on that operation rather than an occasional one.
If the spec comes from somewhere else -- gateways, linters, anything ingesting specs it did not write -- that is a trust boundary, and it becomes remote availability.
If it is your own spec, nothing attacker-controlled starts it, but any client reaching that route then crashes the process. It fires on every call, so normally you would see it the first time that route runs. A spec edited without exercising the route is the case where you would not. That is the difference from GHSA-jpcw, where the attacker supplied the trigger. I have left severity unset; the two cases do not score the same.
On how common it is, I checked rather than guessed. Of roughly 1,500 OpenAPI 3.x specs on apis.guru, one declares content-typed response headers -- VTEX Catalog API -- and not in one corner of it: the shape repeats across its operations, so for that consumer this is not an edge case.
The header names in it (X-AspNet-Version, p3p, a few malformed ones) look like a generator emitting content where it could not infer a schema, and apis.guru only indexes public specs. There is no content-typed response header in this repo's testdata either.
Workarounds
Use schema on the response header, or skip ValidateResponse for those operations. ExcludeResponseBody will not do it -- headers are validated before that check.
Suggested fix
Nil-guard at the top of decodeValue:
if schema == nil {
return nil, false, nil
}
Returning nil leaves a content-typed header unvalidated rather than erroring, so it trades the crash for a silent gap. An error return kills the panic just as well if you would rather it surfaced.
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 0.145.0"
},
"package": {
"ecosystem": "Go",
"name": "github.com/getkin/kin-openapi"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.146.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-100349"
],
"database_specific": {
"cwe_ids": [
"CWE-476"
],
"github_reviewed": true,
"github_reviewed_at": "2026-10-16T19:58:46Z",
"nvd_published_at": "2026-10-15T16:17:30Z",
"severity": "HIGH"
},
"details": "## Summary\n\n`ValidateResponse` panics on a nil pointer when a spec declares a response header with `content` instead of `schema`. `doc.Validate()` accepts that document, so the crash comes from a spec the library just called valid.\n\nSame class as GHSA-jpcw-4wr7-c3vq, which covered the request-parameter case in v0.144.0; this is the response-header path.\n\nAffects `github.com/getkin/kin-openapi` `\u003c= v0.145.0`. Reproduced on master `947498e5a6bd`, Go 1.26.\n\n## Details\n\nA `content`-typed response header has `headerRef.Value.Schema == nil`.\n\n`validate_response.go:181` passes it straight through:\n\n```go\nif decodedValue, found, err = decodeValue(dec, headerName, sm, headerRef.Value.Schema, headerRef.Value.Required); err != nil {\n```\n\nand `req_resp_decoder.go:273` is where it lands:\n\n```go\nif len(schema.Value.AllOf) \u003e 0 {\n```\n\nTwo things I checked rather than assumed. The response does not need to carry the header; declaring it is enough, and then every call for that operation panics. And the body path is fine, `:114` returns early on a nil schema, so this is headers only. The PoC keeps the body case as a control.\n\n## PoC\n\nDrop into `openapi3filter/` and run:\n\n```\ngo test ./openapi3filter/ -run NilSchema -v\n```\n\n```go\npackage openapi3filter_test\n\nimport (\n\t\"context\"\n\t\"io\"\n\t\"net/http\"\n\t\"net/http/httptest\"\n\t\"strings\"\n\t\"testing\"\n\n\t\"github.com/getkin/kin-openapi/openapi3\"\n\t\"github.com/getkin/kin-openapi/openapi3filter\"\n\t\"github.com/getkin/kin-openapi/routers\"\n)\n\nconst specBody = `\nopenapi: 3.0.3\ninfo: {title: poc, version: \"1.0.0\"}\npaths:\n /x:\n get:\n responses:\n \"200\":\n description: ok\n content:\n application/json: {}\n`\n\nconst specHeader = `\nopenapi: 3.0.3\ninfo: {title: poc, version: \"1.0.0\"}\npaths:\n /x:\n get:\n responses:\n \"200\":\n description: ok\n headers:\n X-Thing:\n content:\n application/json: {}\n`\n\nfunc validatedInput(t *testing.T, spec string, hdr http.Header) *openapi3filter.ResponseValidationInput {\n\tt.Helper()\n\tdoc, err := openapi3.NewLoader().LoadFromData([]byte(spec))\n\tif err != nil {\n\t\tt.Fatalf(\"load: %v\", err)\n\t}\n\tif err := doc.Validate(context.Background()); err != nil {\n\t\tt.Fatalf(\"doc.Validate rejected the spec, so this repro proves nothing: %v\", err)\n\t}\n\top := doc.Paths.Find(\"/x\").Get\n\treq := httptest.NewRequest(http.MethodGet, \"/x\", nil)\n\treturn \u0026openapi3filter.ResponseValidationInput{\n\t\tRequestValidationInput: \u0026openapi3filter.RequestValidationInput{\n\t\t\tRequest: req,\n\t\t\tRoute: \u0026routers.Route{Spec: doc, Operation: op, Method: http.MethodGet, Path: \"/x\"},\n\t\t},\n\t\tStatus: 200,\n\t\tHeader: hdr,\n\t\tBody: io.NopCloser(strings.NewReader(`{}`)),\n\t}\n}\n\n// Control: body path returns early on a nil schema.\nfunc TestControl_ResponseBodyNilSchema(t *testing.T) {\n\tin := validatedInput(t, specBody, http.Header{\"Content-Type\": {\"application/json\"}})\n\tdefer func() {\n\t\tif r := recover(); r != nil {\n\t\t\tt.Errorf(\"body path panicked, expected clean return: %v\", r)\n\t\t}\n\t}()\n\tif err := openapi3filter.ValidateResponse(context.Background(), in); err != nil {\n\t\tt.Errorf(\"body path returned %v, expected nil\", err)\n\t}\n}\n\n// The response does not carry X-Thing. Declaring the header is enough.\nfunc TestResponseHeaderNilSchema(t *testing.T) {\n\tin := validatedInput(t, specHeader, http.Header{\"Content-Type\": {\"application/json\"}})\n\tif err := openapi3filter.ValidateResponse(context.Background(), in); err != nil {\n\t\tt.Fatalf(\"no panic; got err = %v (already fixed?)\", err)\n\t}\n}\n```\n\nOutput:\n\n```\n=== RUN TestControl_ResponseBodyNilSchema\n--- PASS: TestControl_ResponseBodyNilSchema (0.00s)\n=== RUN TestResponseHeaderNilSchema\n--- FAIL: TestResponseHeaderNilSchema (0.00s)\npanic: runtime error: invalid memory address or nil pointer dereference [recovered, repanicked]\n[signal SIGSEGV: segmentation violation code=0x1 addr=0x20 pc=0x7f6d2a]\n\ngithub.com/getkin/kin-openapi/openapi3filter.decodeValue(...)\n\t/src/openapi3filter/req_resp_decoder.go:273 +0x4a\ngithub.com/getkin/kin-openapi/openapi3filter.validateResponseHeader(...)\n\t/src/openapi3filter/validate_response.go:181 +0x146\ngithub.com/getkin/kin-openapi/openapi3filter.ValidateResponse(...)\n\t/src/openapi3filter/validate_response.go:89 +0xafa\n```\n\nThe `FAIL` is the point: control passes, the header case takes the binary down.\n\nRan in a container against a fresh clone of master. I have not tried it against a live proxy.\n\n## Impact\n\nA process crash instead of an error return. Declaring the header is enough, so it is every call on that operation rather than an occasional one.\n\nIf the spec comes from somewhere else -- gateways, linters, anything ingesting specs it did not write -- that is a trust boundary, and it becomes remote availability.\n\nIf it is your own spec, nothing attacker-controlled starts it, but any client reaching that route then crashes the process. It fires on every call, so normally you would see it the first time that route runs. A spec edited without exercising the route is the case where you would not. That is the difference from GHSA-jpcw, where the attacker supplied the trigger. I have left severity unset; the two cases do not score the same.\n\nOn how common it is, I checked rather than guessed. Of roughly 1,500 OpenAPI 3.x specs on apis.guru, one declares content-typed response headers -- VTEX Catalog API -- and not in one corner of it: the shape repeats across its operations, so for that consumer this is not an edge case.\n\nThe header names in it (`X-AspNet-Version`, `p3p`, a few malformed ones) look like a generator emitting `content` where it could not infer a schema, and apis.guru only indexes public specs. There is no content-typed response header in this repo\u0027s testdata either.\n\n## Workarounds\n\nUse `schema` on the response header, or skip `ValidateResponse` for those operations. `ExcludeResponseBody` will not do it -- headers are validated before that check.\n\n## Suggested fix\n\nNil-guard at the top of `decodeValue`:\n\n```go\nif schema == nil {\n\treturn nil, false, nil\n}\n```\n\nReturning nil leaves a `content`-typed header unvalidated rather than erroring, so it trades the crash for a silent gap. An error return kills the panic just as well if you would rather it surfaced.",
"id": "GHSA-74vm-87hj-r66f",
"modified": "2026-10-16T19:58:46Z",
"published": "2026-10-16T19:58:46Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/getkin/kin-openapi/security/advisories/GHSA-74vm-87hj-r66f"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-100349"
},
{
"type": "WEB",
"url": "https://github.com/getkin/kin-openapi/commit/f5441d67f855ed34d975b3b85665764644ab1459"
},
{
"type": "PACKAGE",
"url": "https://github.com/getkin/kin-openapi"
},
{
"type": "WEB",
"url": "https://github.com/getkin/kin-openapi/releases/tag/v0.146.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "kin-openapi openapi3filter: nil-pointer panic in ValidateResponse on a response header declared with content"
}
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
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.