CWE-770
AllowedAllocation of Resources Without Limits or Throttling
Abstraction: Base · Status: Incomplete
The product allocates a reusable resource or group of resources on behalf of an actor without imposing any intended restrictions on the size or number of resources that can be allocated.
3707 vulnerabilities reference this CWE, most recent first.
GHSA-H9FH-4377-WFQ4
Vulnerability from github – Published: 2024-07-05 09:33 – Updated: 2025-11-04 00:30In the Linux kernel, the following vulnerability has been resolved:
mm/vmalloc: fix vmalloc which may return null if called with __GFP_NOFAIL
commit a421ef303008 ("mm: allow !GFP_KERNEL allocations for kvmalloc") includes support for __GFP_NOFAIL, but it presents a conflict with commit dd544141b9eb ("vmalloc: back off when the current task is OOM-killed"). A possible scenario is as follows:
process-a __vmalloc_node_range(GFP_KERNEL | __GFP_NOFAIL) __vmalloc_area_node() vm_area_alloc_pages() --> oom-killer send SIGKILL to process-a if (fatal_signal_pending(current)) break; --> return NULL;
To fix this, do not check fatal_signal_pending() in vm_area_alloc_pages() if __GFP_NOFAIL set.
This issue occurred during OPLUS KASAN TEST. Below is part of the log -> oom-killer sends signal to process [65731.222840] [ T1308] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/apps/uid_10198,task=gs.intelligence,pid=32454,uid=10198
[65731.259685] [T32454] Call trace: [65731.259698] [T32454] dump_backtrace+0xf4/0x118 [65731.259734] [T32454] show_stack+0x18/0x24 [65731.259756] [T32454] dump_stack_lvl+0x60/0x7c [65731.259781] [T32454] dump_stack+0x18/0x38 [65731.259800] [T32454] mrdump_common_die+0x250/0x39c [mrdump] [65731.259936] [T32454] ipanic_die+0x20/0x34 [mrdump] [65731.260019] [T32454] atomic_notifier_call_chain+0xb4/0xfc [65731.260047] [T32454] notify_die+0x114/0x198 [65731.260073] [T32454] die+0xf4/0x5b4 [65731.260098] [T32454] die_kernel_fault+0x80/0x98 [65731.260124] [T32454] __do_kernel_fault+0x160/0x2a8 [65731.260146] [T32454] do_bad_area+0x68/0x148 [65731.260174] [T32454] do_mem_abort+0x151c/0x1b34 [65731.260204] [T32454] el1_abort+0x3c/0x5c [65731.260227] [T32454] el1h_64_sync_handler+0x54/0x90 [65731.260248] [T32454] el1h_64_sync+0x68/0x6c
[65731.260269] [T32454] z_erofs_decompress_queue+0x7f0/0x2258 --> be->decompressed_pages = kvcalloc(be->nr_pages, sizeof(struct page *), GFP_KERNEL | __GFP_NOFAIL); kernel panic by NULL pointer dereference. erofs assume kvmalloc with __GFP_NOFAIL never return NULL. [65731.260293] [T32454] z_erofs_runqueue+0xf30/0x104c [65731.260314] [T32454] z_erofs_readahead+0x4f0/0x968 [65731.260339] [T32454] read_pages+0x170/0xadc [65731.260364] [T32454] page_cache_ra_unbounded+0x874/0xf30 [65731.260388] [T32454] page_cache_ra_order+0x24c/0x714 [65731.260411] [T32454] filemap_fault+0xbf0/0x1a74 [65731.260437] [T32454] __do_fault+0xd0/0x33c [65731.260462] [T32454] handle_mm_fault+0xf74/0x3fe0 [65731.260486] [T32454] do_mem_abort+0x54c/0x1b34 [65731.260509] [T32454] el0_da+0x44/0x94 [65731.260531] [T32454] el0t_64_sync_handler+0x98/0xb4 [65731.260553] [T32454] el0t_64_sync+0x198/0x19c
{
"affected": [],
"aliases": [
"CVE-2024-39474"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": false,
"github_reviewed_at": null,
"nvd_published_at": "2024-07-05T07:15:10Z",
"severity": "MODERATE"
},
"details": "In the Linux kernel, the following vulnerability has been resolved:\n\nmm/vmalloc: fix vmalloc which may return null if called with __GFP_NOFAIL\n\ncommit a421ef303008 (\"mm: allow !GFP_KERNEL allocations for kvmalloc\")\nincludes support for __GFP_NOFAIL, but it presents a conflict with commit\ndd544141b9eb (\"vmalloc: back off when the current task is OOM-killed\"). A\npossible scenario is as follows:\n\nprocess-a\n__vmalloc_node_range(GFP_KERNEL | __GFP_NOFAIL)\n __vmalloc_area_node()\n vm_area_alloc_pages()\n\t\t--\u003e oom-killer send SIGKILL to process-a\n if (fatal_signal_pending(current)) break;\n--\u003e return NULL;\n\nTo fix this, do not check fatal_signal_pending() in vm_area_alloc_pages()\nif __GFP_NOFAIL set.\n\nThis issue occurred during OPLUS KASAN TEST. Below is part of the log\n-\u003e oom-killer sends signal to process\n[65731.222840] [ T1308] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/apps/uid_10198,task=gs.intelligence,pid=32454,uid=10198\n\n[65731.259685] [T32454] Call trace:\n[65731.259698] [T32454] dump_backtrace+0xf4/0x118\n[65731.259734] [T32454] show_stack+0x18/0x24\n[65731.259756] [T32454] dump_stack_lvl+0x60/0x7c\n[65731.259781] [T32454] dump_stack+0x18/0x38\n[65731.259800] [T32454] mrdump_common_die+0x250/0x39c [mrdump]\n[65731.259936] [T32454] ipanic_die+0x20/0x34 [mrdump]\n[65731.260019] [T32454] atomic_notifier_call_chain+0xb4/0xfc\n[65731.260047] [T32454] notify_die+0x114/0x198\n[65731.260073] [T32454] die+0xf4/0x5b4\n[65731.260098] [T32454] die_kernel_fault+0x80/0x98\n[65731.260124] [T32454] __do_kernel_fault+0x160/0x2a8\n[65731.260146] [T32454] do_bad_area+0x68/0x148\n[65731.260174] [T32454] do_mem_abort+0x151c/0x1b34\n[65731.260204] [T32454] el1_abort+0x3c/0x5c\n[65731.260227] [T32454] el1h_64_sync_handler+0x54/0x90\n[65731.260248] [T32454] el1h_64_sync+0x68/0x6c\n\n[65731.260269] [T32454] z_erofs_decompress_queue+0x7f0/0x2258\n--\u003e be-\u003edecompressed_pages = kvcalloc(be-\u003enr_pages, sizeof(struct page *), GFP_KERNEL | __GFP_NOFAIL);\n\tkernel panic by NULL pointer dereference.\n\terofs assume kvmalloc with __GFP_NOFAIL never return NULL.\n[65731.260293] [T32454] z_erofs_runqueue+0xf30/0x104c\n[65731.260314] [T32454] z_erofs_readahead+0x4f0/0x968\n[65731.260339] [T32454] read_pages+0x170/0xadc\n[65731.260364] [T32454] page_cache_ra_unbounded+0x874/0xf30\n[65731.260388] [T32454] page_cache_ra_order+0x24c/0x714\n[65731.260411] [T32454] filemap_fault+0xbf0/0x1a74\n[65731.260437] [T32454] __do_fault+0xd0/0x33c\n[65731.260462] [T32454] handle_mm_fault+0xf74/0x3fe0\n[65731.260486] [T32454] do_mem_abort+0x54c/0x1b34\n[65731.260509] [T32454] el0_da+0x44/0x94\n[65731.260531] [T32454] el0t_64_sync_handler+0x98/0xb4\n[65731.260553] [T32454] el0t_64_sync+0x198/0x19c",
"id": "GHSA-h9fh-4377-wfq4",
"modified": "2025-11-04T00:30:50Z",
"published": "2024-07-05T09:33:44Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2024-39474"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/198a80833e3421d4c9820a4ae907120adf598c91"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/758678b65164b2158fc1de411092191cb3c394d4"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/8e0545c83d672750632f46e3f9ad95c48c91a0fc"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/c55d3564ad25ce87ab7cc6af251f9574faebd8da"
},
{
"type": "WEB",
"url": "https://lists.debian.org/debian-lts-announce/2025/01/msg00001.html"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
]
}
GHSA-H9HM-M2XJ-4RQ9
Vulnerability from github – Published: 2026-05-08 19:12 – Updated: 2026-05-08 19:12Summary
A composite denial-of-service vulnerability in Zebra's block discovery pipeline allows an unauthenticated remote attacker to permanently halt all new block discovery on a targeted node. The attack exploits three independent weaknesses in the gossip, syncer, and download subsystems — all exercisable from a single TCP connection — to create a monotonically growing block deficit that never self-heals.
Severity
Critical — This is a Denial of Service vulnerability that requires no authentication, no special privileges, and only a single peer connection. The halt is permanent: the node will never recover without operator intervention.
Affected Versions
All Zebra versions prior to 4.4.0.
Description
Zebra discovers new blocks through two complementary paths: a gossip path (peers announce blocks via inv messages, triggering individual block downloads) and a syncer path (Zebra periodically queries peers with FindBlocks/FindHeaders to discover chains of missing blocks). Both paths must function for normal operation.
The gossip path was vulnerable because there was no per-connection rate limit on inv messages. A single connection could send enough sequential inv messages with fake block hashes to fill the entire gossip download queue in under a millisecond. The FullQueue return value was silently ignored, so legitimate block announcements from honest peers were dropped with no warning.
The syncer backup path could be degraded by responding with empty inv to FindBlocks requests and with NotFound to block download requests. Both are valid protocol responses that carried zero misbehavior penalty. The attacker's connection was never banned and never disconnected, allowing the degradation to persist indefinitely.
Combining these two vectors, an attacker could suppress both block discovery paths simultaneously from a single connection, causing the node to fall permanently behind the chain tip.
Impact
Denial of Service
- Attack Vector: Network, unauthenticated. Requires only a single TCP peer connection.
- Effect: Permanent halt of block discovery. The targeted node falls behind the chain tip and never recovers without operator intervention.
- Scope: Any Zebra node reachable by the attacker over the peer-to-peer network.
Fixed Versions
This issue is fixed in Zebra 4.4.0.
The fix drops connections that send empty responses to FindBlocks and FindHeaders messages, preventing attackers from degrading the syncer path without consequence.
Mitigation
Users should upgrade to Zebra 4.4.0 or later immediately.
There are no known workarounds for this issue. Immediate upgrade is the only way to protect against this attack.
Credits
Zebra the researcher who reported this issue through the coordinated disclosure process.
{
"affected": [
{
"package": {
"ecosystem": "crates.io",
"name": "zebrad"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "4.4.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-44499"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": true,
"github_reviewed_at": "2026-05-08T19:12:35Z",
"nvd_published_at": "2026-05-08T16:16:13Z",
"severity": "HIGH"
},
"details": "## Summary\n\nA composite denial-of-service vulnerability in Zebra\u0027s block discovery pipeline allows an unauthenticated remote attacker to permanently halt all new block discovery on a targeted node. The attack exploits three independent weaknesses in the gossip, syncer, and download subsystems \u2014 all exercisable from a single TCP connection \u2014 to create a monotonically growing block deficit that never self-heals.\n\n## Severity\n\n**Critical** \u2014 This is a Denial of Service vulnerability that requires no authentication, no special privileges, and only a single peer connection. The halt is permanent: the node will never recover without operator intervention.\n\n## Affected Versions\n\nAll Zebra versions prior to 4.4.0.\n\n## Description\n\nZebra discovers new blocks through two complementary paths: a gossip path (peers announce blocks via `inv` messages, triggering individual block downloads) and a syncer path (Zebra periodically queries peers with `FindBlocks`/`FindHeaders` to discover chains of missing blocks). Both paths must function for normal operation.\n\nThe gossip path was vulnerable because there was no per-connection rate limit on `inv` messages. A single connection could send enough sequential `inv` messages with fake block hashes to fill the entire gossip download queue in under a millisecond. The `FullQueue` return value was silently ignored, so legitimate block announcements from honest peers were dropped with no warning.\n\nThe syncer backup path could be degraded by responding with empty `inv` to `FindBlocks` requests and with `NotFound` to block download requests. Both are valid protocol responses that carried zero misbehavior penalty. The attacker\u0027s connection was never banned and never disconnected, allowing the degradation to persist indefinitely.\n\nCombining these two vectors, an attacker could suppress both block discovery paths simultaneously from a single connection, causing the node to fall permanently behind the chain tip.\n\n## Impact\n\n**Denial of Service**\n\n* **Attack Vector:** Network, unauthenticated. Requires only a single TCP peer connection.\n* **Effect:** Permanent halt of block discovery. The targeted node falls behind the chain tip and never recovers without operator intervention.\n* **Scope:** Any Zebra node reachable by the attacker over the peer-to-peer network.\n\n## Fixed Versions\n\nThis issue is fixed in Zebra 4.4.0.\n\nThe fix drops connections that send empty responses to `FindBlocks` and `FindHeaders` messages, preventing attackers from degrading the syncer path without consequence.\n\n## Mitigation\n\nUsers should upgrade to Zebra 4.4.0 or later immediately.\n\nThere are no known workarounds for this issue. Immediate upgrade is the only way to protect against this attack.\n\n## Credits\n\nZebra the researcher who reported this issue through the coordinated disclosure process.",
"id": "GHSA-h9hm-m2xj-4rq9",
"modified": "2026-05-08T19:12:35Z",
"published": "2026-05-08T19:12:35Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/ZcashFoundation/zebra/security/advisories/GHSA-h9hm-m2xj-4rq9"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-44499"
},
{
"type": "PACKAGE",
"url": "https://github.com/ZcashFoundation/zebra"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "Zebra has Permanent Block Discovery Halt via Gossip Queue Saturation and Syncer Poisoning"
}
GHSA-H9MQ-F6Q5-6C8M
Vulnerability from github – Published: 2024-07-30 09:31 – Updated: 2024-07-30 15:08GraphQL Java (aka graphql-java) before 21.5 does not properly consider ExecutableNormalizedFields (ENFs) as part of preventing denial of service via introspection queries. 20.9 and 19.11 are also fixed versions.
{
"affected": [
{
"package": {
"ecosystem": "Maven",
"name": "com.graphql-java:graphql-java"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "19.11"
}
],
"type": "ECOSYSTEM"
}
]
},
{
"package": {
"ecosystem": "Maven",
"name": "com.graphql-java:graphql-java"
},
"ranges": [
{
"events": [
{
"introduced": "20.0"
},
{
"fixed": "20.9"
}
],
"type": "ECOSYSTEM"
}
]
},
{
"package": {
"ecosystem": "Maven",
"name": "com.graphql-java:graphql-java"
},
"ranges": [
{
"events": [
{
"introduced": "21.0"
},
{
"fixed": "21.5"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2024-40094"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": true,
"github_reviewed_at": "2024-07-30T15:08:48Z",
"nvd_published_at": "2024-07-30T07:15:01Z",
"severity": "HIGH"
},
"details": "GraphQL Java (aka graphql-java) before 21.5 does not properly consider ExecutableNormalizedFields (ENFs) as part of preventing denial of service via introspection queries. 20.9 and 19.11 are also fixed versions.",
"id": "GHSA-h9mq-f6q5-6c8m",
"modified": "2024-07-30T15:08:48Z",
"published": "2024-07-30T09:31:50Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2024-40094"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/pull/3539"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/commit/16c159111507ef04d7e1839b2c23281d90c42b2b"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/commit/469caf6ee600ab6709ad5e8a06f371fe2ef3b8dd"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/commit/97743bc1b5caa2b0bd894dc8e128b47e4d771e4a"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/commit/fc6f304e66cab18b6d06a80c7009524938939a03"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/discussions/3641"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/releases/tag/v19.11"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/releases/tag/v20.9"
},
{
"type": "WEB",
"url": "https://github.com/graphql-java/graphql-java/releases/tag/v21.5"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
},
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "GraphQL Java does not properly consider ExecutableNormalizedFields (ENFs) as part of preventing denial of service"
}
GHSA-H9MR-9VH5-65GW
Vulnerability from github – Published: 2026-08-06 12:31 – Updated: 2026-08-07 00:31Apache CXF allows to control the maximum attachment size via the "attachment-max-size". Prior to Apache CXF 4.2.3 and 4.1.8 and 3.6.12, there was no default placed on this size, meaning that a denial of service attack is possible if the user doesn't explicitly set the limit. Users should update to Apache CXF 4.2.3 or 4.1.8 or 3.6.12 which fixes this problem by imposing a default attachment size limit of 50mb.
{
"affected": [],
"aliases": [
"CVE-2026-54225"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": false,
"github_reviewed_at": null,
"nvd_published_at": "2026-08-06T11:16:30Z",
"severity": "HIGH"
},
"details": "Apache CXF allows to control the maximum attachment size via the\u00a0\"attachment-max-size\". Prior to Apache CXF 4.2.3 and 4.1.8 and 3.6.12, there was no default placed on this size, meaning that a denial of service attack is possible if the user doesn\u0027t explicitly set the limit. Users should update to Apache CXF 4.2.3 or 4.1.8 or 3.6.12 which fixes this problem by imposing a default attachment size limit of 50mb.",
"id": "GHSA-h9mr-9vh5-65gw",
"modified": "2026-08-07T00:31:09Z",
"published": "2026-08-06T12:31:13Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-54225"
},
{
"type": "WEB",
"url": "https://lists.apache.org/thread/h2bjqm6g58z0j6893qzh728kdtk1byfy"
},
{
"type": "WEB",
"url": "http://www.openwall.com/lists/oss-security/2026/08/06/14"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
]
}
GHSA-H9MW-H4QC-F5JF
Vulnerability from github – Published: 2026-04-08 15:05 – Updated: 2026-04-08 15:05CVSS 6.5 Medium — The GraphQL API served by kubernetes-graphql-gateway is vulnerable to Denial-of-Service (DoS) attacks due to a complete absence of query resource controls (depth limiting, complexity analysis, response size capping, and rate limiting). An authenticated attacker can craft queries that force the server to compute and serialize multi-megabyte responses, consuming significant CPU, memory, and network bandwidth. Repeated requests can exhaust server resources and degrade or deny service to legitimate users.
Note: A previous version of this advisory (based on pre-v1 code) documented an unauthenticated attack surface via an HTTP GET method bypass in the former
registry.go. That bypass has been removed in v1 — all requests now require a Bearer token. The CVSS score has been adjusted from 7.5 to 6.5 accordingly (Privileges Required: None → Low). CWE-306 (Missing Authentication for Critical Function) no longer applies.
Root Cause
The kubernetes-graphql-gateway uses the graphql-go/graphql library (v0.8.1) with the graphql-go/handler HTTP handler. The handler is instantiated in gateway/gateway/graphql/graphql.go with only cosmetic configuration — no resource limits:
// gateway/gateway/graphql/graphql.go — CreateHandler()
func (s *GraphQLServer) CreateHandler(schema *graphql.Schema) *GraphQLHandler {
graphqlHandler := handler.New(&handler.Config{
Schema: schema,
Pretty: s.config.Pretty,
Playground: s.config.Playground,
GraphiQL: s.config.GraphiQL,
})
return &GraphQLHandler{
Schema: schema,
Handler: graphqlHandler,
}
}
The handler.Config struct does not include MaxDepth, MaxComplexity, MaxResponseSize, or any equivalent fields. Neither the graphql-go/handler nor the underlying graphql-go/graphql library provides built-in query depth or complexity analysis.
The application configuration (gateway/gateway/config/config.go) has no fields for resource limits:
// gateway/gateway/config/config.go — GraphQL config
type GraphQL struct {
Pretty bool
Playground bool
GraphiQL bool
}
No rate limiting, throttling, or request size controls exist anywhere in the codebase.
Authentication Model
All requests pass through the HTTP handler in gateway/http/http.go, which extracts a Bearer token and injects it into the request context:
// gateway/http/http.go — Token extraction (applied to all methods)
s.Handle("/api/clusters/{clusterName}", http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
clusterName := r.PathValue("clusterName")
authHeader := r.Header.Get("Authorization")
token := strings.TrimPrefix(authHeader, "Bearer ")
ctx := utilscontext.SetToken(r.Context(), token)
ctx = utilscontext.SetCluster(ctx, clusterName)
c.Gateway.ServeHTTP(w, r.WithContext(ctx))
}))
The token is enforced at the Kubernetes API layer via gateway/gateway/roundtripper/bearer.go, which returns HTTP 401 for requests without a valid token. However, the GraphQL execution engine (query parsing, schema validation, introspection) still runs before the Kubernetes API is contacted — meaning authenticated users can trigger expensive operations that consume server resources without hitting the K8s API at all.
Attack Vectors
1. Nested Introspection Field Expansion
The GraphQL schema contains 3,508 types (Kubernetes resources + platform CRDs). Introspection meta-fields (__schema, __type) allow recursive field expansion. Each additional nesting level multiplies the response size exponentially. A single full introspection query generates ~5.2 MB of response data in ~1.15s.
2. Parallel Request Amplification
Without rate limiting, an authenticated attacker can issue many concurrent expensive queries. 5 parallel requests generate ~18.6 MB total response in under 4 seconds with no throttling. At scale (e.g. 999 concurrent requests), the backend becomes unresponsive and returns 503 to all users.
3. Subscription Resource Exhaustion
The HandleSubscription() method in gateway/gateway/graphql/graphql.go processes SSE (Server-Sent Events) subscription requests. A malicious authenticated client can open many subscription channels simultaneously, holding server connections and memory indefinitely:
// gateway/gateway/graphql/graphql.go — HandleSubscription()
subscriptionChannel := graphql.Subscribe(subscriptionParams)
for res := range subscriptionChannel {
// ... marshal and flush indefinitely ...
}
There is no limit on the number of concurrent subscriptions, no idle timeout, and no per-client connection cap.
4. Deep Query Execution
Authenticated users can submit arbitrarily deep and complex GraphQL queries. The GraphQL execution engine processes the full query — consuming CPU and memory for schema validation, field resolution, and error/response formatting — before any Kubernetes API authorization is checked. The request handling in gateway/gateway/endpoint/endpoint.go passes directly to the handler with no query guards:
// gateway/gateway/endpoint/endpoint.go — ServeHTTP()
func (e *Endpoint) ServeHTTP(w http.ResponseWriter, r *http.Request) {
if e.handler == nil || e.handler.Handler == nil {
http.Error(w, "Endpoint not ready", http.StatusServiceUnavailable)
return
}
if r.Header.Get("Accept") == "text/event-stream" {
e.graphqlServer.HandleSubscription(w, r, e.handler.Schema)
return
}
e.handler.Handler.ServeHTTP(w, r)
}
Impact
- Availability (High): Service denial achievable — concurrent expensive queries cause backend to become unresponsive (503 for all users). With 3,508 types and no depth limits, each introspection query generates a ~5.2 MB response. The absence of rate limiting, query complexity controls, and response size caps allows an authenticated attacker to exhaust server CPU, memory, and bandwidth.
- Confidentiality (None): Information disclosure is covered in a separate finding.
- Integrity (None): No data modification possible.
Affected Components
gateway/gateway/graphql/graphql.go— Handler creation with no resource limits; subscription handler with no connection limitsgateway/gateway/endpoint/endpoint.go— Direct passthrough to handler, no query depth/complexity middlewaregateway/gateway/config/config.go— No configuration fields for resource limitsgateway/http/http.go— No rate limiting middlewaregraphql-go/graphqllibrary — No built-in depth/complexity limitinggraphql-go/handler— No resource limit configuration options
Recommendations
- Disable Introspection in Production — As a defense-in-depth measure, disable introspection in non-development environments. This removes the highest-cost query path. If GraphiQL/Playground must remain accessible for development, gate it behind an environment flag.
- Implement Query Depth and Complexity Limiting — Implement middleware that parses the query AST and rejects queries exceeding configurable thresholds before execution. Recommended maximum depth: 10 levels. Assign cost values to fields and enforce a maximum query cost budget — introspection meta-fields (
__schema,__type) should carry elevated costs. Alternatively, consider migrating to a GraphQL library with built-in depth/complexity support (e.g.,gqlgenwith its complexity extension, orgraph-gophers/graphql-gowith itsMaxDepthoption). - Implement Rate Limiting and Response Size Controls — Add per-user rate limiting on the GraphQL endpoint. Suggested thresholds: 60 requests/minute for authenticated users, 2 requests/minute for introspection queries. Cap response payload size (e.g., 5 MB). For subscriptions, enforce maximum concurrent connections per client, idle timeouts, and maximum subscription duration.
- Add Resource Limit Configuration — Extend the
GraphQLstruct ingateway/gateway/config/config.goto expose all resource limits (max query depth, max complexity, max response size, rate limit thresholds) as configurable parameters. This ensures all protective thresholds can be tuned per environment without code changes.
References
- OWASP GraphQL Cheat Sheet — Resource Limits
- OWASP API4:2023 — Unrestricted Resource Consumption
- CWE-770: Allocation of Resources Without Limits or Throttling
- CWE-400: Uncontrolled Resource Consumption
Classification
- CWE-770 — Allocation of Resources Without Limits or Throttling
- CWE-400 — Uncontrolled Resource Consumption
- OWASP Top 10 2021: A05:2021 — Security Misconfiguration
- OWASP API Security Top 10: API4:2023 — Unrestricted Resource Consumption
- STRIDE: Denial of Service (D)
Internal Reference
HASI2026141-32 — Due: 2026-04-16
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 1.2.8"
},
"package": {
"ecosystem": "Go",
"name": "github.com/platform-mesh/kubernetes-graphql-gateway"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "1.2.9"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [],
"database_specific": {
"cwe_ids": [
"CWE-400",
"CWE-770"
],
"github_reviewed": true,
"github_reviewed_at": "2026-04-08T15:05:10Z",
"nvd_published_at": null,
"severity": "MODERATE"
},
"details": "**CVSS 6.5 Medium** \u2014 The GraphQL API served by kubernetes-graphql-gateway is vulnerable to Denial-of-Service (DoS) attacks due to a complete absence of query resource controls (depth limiting, complexity analysis, response size capping, and rate limiting). An authenticated attacker can craft queries that force the server to compute and serialize multi-megabyte responses, consuming significant CPU, memory, and network bandwidth. Repeated requests can exhaust server resources and degrade or deny service to legitimate users.\n\n\u003e **Note:** A previous version of this advisory (based on pre-v1 code) documented an unauthenticated attack surface via an HTTP GET method bypass in the former `registry.go`. That bypass has been removed in v1 \u2014 all requests now require a Bearer token. The CVSS score has been adjusted from 7.5 to 6.5 accordingly (Privileges Required: None \u2192 Low). CWE-306 (Missing Authentication for Critical Function) no longer applies.\n\n## Root Cause\n\nThe kubernetes-graphql-gateway uses the `graphql-go/graphql` library (v0.8.1) with the `graphql-go/handler` HTTP handler. The handler is instantiated in `gateway/gateway/graphql/graphql.go` with only cosmetic configuration \u2014 no resource limits:\n\n```go\n// gateway/gateway/graphql/graphql.go \u2014 CreateHandler()\nfunc (s *GraphQLServer) CreateHandler(schema *graphql.Schema) *GraphQLHandler {\n graphqlHandler := handler.New(\u0026handler.Config{\n Schema: schema,\n Pretty: s.config.Pretty,\n Playground: s.config.Playground,\n GraphiQL: s.config.GraphiQL,\n })\n return \u0026GraphQLHandler{\n Schema: schema,\n Handler: graphqlHandler,\n }\n}\n```\n\nThe `handler.Config` struct does not include `MaxDepth`, `MaxComplexity`, `MaxResponseSize`, or any equivalent fields. Neither the `graphql-go/handler` nor the underlying `graphql-go/graphql` library provides built-in query depth or complexity analysis.\n\nThe application configuration (`gateway/gateway/config/config.go`) has no fields for resource limits:\n\n```go\n// gateway/gateway/config/config.go \u2014 GraphQL config\ntype GraphQL struct {\n Pretty bool\n Playground bool\n GraphiQL bool\n}\n```\n\nNo rate limiting, throttling, or request size controls exist anywhere in the codebase.\n\n## Authentication Model\n\nAll requests pass through the HTTP handler in `gateway/http/http.go`, which extracts a Bearer token and injects it into the request context:\n\n```go\n// gateway/http/http.go \u2014 Token extraction (applied to all methods)\ns.Handle(\"/api/clusters/{clusterName}\", http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {\n clusterName := r.PathValue(\"clusterName\")\n authHeader := r.Header.Get(\"Authorization\")\n token := strings.TrimPrefix(authHeader, \"Bearer \")\n ctx := utilscontext.SetToken(r.Context(), token)\n ctx = utilscontext.SetCluster(ctx, clusterName)\n c.Gateway.ServeHTTP(w, r.WithContext(ctx))\n}))\n```\n\nThe token is enforced at the Kubernetes API layer via `gateway/gateway/roundtripper/bearer.go`, which returns HTTP 401 for requests without a valid token. However, the GraphQL execution engine (query parsing, schema validation, introspection) still runs **before** the Kubernetes API is contacted \u2014 meaning authenticated users can trigger expensive operations that consume server resources without hitting the K8s API at all.\n\n## Attack Vectors\n\n### 1. Nested Introspection Field Expansion\n\nThe GraphQL schema contains 3,508 types (Kubernetes resources + platform CRDs). Introspection meta-fields (`__schema`, `__type`) allow recursive field expansion. Each additional nesting level multiplies the response size exponentially. A single full introspection query generates ~5.2 MB of response data in ~1.15s.\n\n### 2. Parallel Request Amplification\n\nWithout rate limiting, an authenticated attacker can issue many concurrent expensive queries. 5 parallel requests generate ~18.6 MB total response in under 4 seconds with no throttling. At scale (e.g. 999 concurrent requests), the backend becomes unresponsive and returns 503 to all users.\n\n### 3. Subscription Resource Exhaustion\n\nThe `HandleSubscription()` method in `gateway/gateway/graphql/graphql.go` processes SSE (Server-Sent Events) subscription requests. A malicious authenticated client can open many subscription channels simultaneously, holding server connections and memory indefinitely:\n\n```go\n// gateway/gateway/graphql/graphql.go \u2014 HandleSubscription()\nsubscriptionChannel := graphql.Subscribe(subscriptionParams)\nfor res := range subscriptionChannel {\n // ... marshal and flush indefinitely ...\n}\n```\n\nThere is no limit on the number of concurrent subscriptions, no idle timeout, and no per-client connection cap.\n\n### 4. Deep Query Execution\n\nAuthenticated users can submit arbitrarily deep and complex GraphQL queries. The GraphQL execution engine processes the full query \u2014 consuming CPU and memory for schema validation, field resolution, and error/response formatting \u2014 before any Kubernetes API authorization is checked. The request handling in `gateway/gateway/endpoint/endpoint.go` passes directly to the handler with no query guards:\n\n```go\n// gateway/gateway/endpoint/endpoint.go \u2014 ServeHTTP()\nfunc (e *Endpoint) ServeHTTP(w http.ResponseWriter, r *http.Request) {\n if e.handler == nil || e.handler.Handler == nil {\n http.Error(w, \"Endpoint not ready\", http.StatusServiceUnavailable)\n return\n }\n if r.Header.Get(\"Accept\") == \"text/event-stream\" {\n e.graphqlServer.HandleSubscription(w, r, e.handler.Schema)\n return\n }\n e.handler.Handler.ServeHTTP(w, r)\n}\n```\n\n## Impact\n\n- **Availability (High):** Service denial achievable \u2014 concurrent expensive queries cause backend to become unresponsive (503 for all users). With 3,508 types and no depth limits, each introspection query generates a ~5.2 MB response. The absence of rate limiting, query complexity controls, and response size caps allows an authenticated attacker to exhaust server CPU, memory, and bandwidth.\n- **Confidentiality (None):** Information disclosure is covered in a separate finding.\n- **Integrity (None):** No data modification possible.\n\n## Affected Components\n\n- `gateway/gateway/graphql/graphql.go` \u2014 Handler creation with no resource limits; subscription handler with no connection limits\n- `gateway/gateway/endpoint/endpoint.go` \u2014 Direct passthrough to handler, no query depth/complexity middleware\n- `gateway/gateway/config/config.go` \u2014 No configuration fields for resource limits\n- `gateway/http/http.go` \u2014 No rate limiting middleware\n- `graphql-go/graphql` library \u2014 No built-in depth/complexity limiting\n- `graphql-go/handler` \u2014 No resource limit configuration options\n\n## Recommendations\n\n1. **Disable Introspection in Production** \u2014 As a defense-in-depth measure, disable introspection in non-development environments. This removes the highest-cost query path. If GraphiQL/Playground must remain accessible for development, gate it behind an environment flag.\n2. **Implement Query Depth and Complexity Limiting** \u2014 Implement middleware that parses the query AST and rejects queries exceeding configurable thresholds before execution. Recommended maximum depth: 10 levels. Assign cost values to fields and enforce a maximum query cost budget \u2014 introspection meta-fields (`__schema`, `__type`) should carry elevated costs. Alternatively, consider migrating to a GraphQL library with built-in depth/complexity support (e.g., `gqlgen` with its complexity extension, or `graph-gophers/graphql-go` with its `MaxDepth` option).\n3. **Implement Rate Limiting and Response Size Controls** \u2014 Add per-user rate limiting on the GraphQL endpoint. Suggested thresholds: 60 requests/minute for authenticated users, 2 requests/minute for introspection queries. Cap response payload size (e.g., 5 MB). For subscriptions, enforce maximum concurrent connections per client, idle timeouts, and maximum subscription duration.\n4. **Add Resource Limit Configuration** \u2014 Extend the `GraphQL` struct in `gateway/gateway/config/config.go` to expose all resource limits (max query depth, max complexity, max response size, rate limit thresholds) as configurable parameters. This ensures all protective thresholds can be tuned per environment without code changes.\n\n## References\n\n- [OWASP GraphQL Cheat Sheet \u2014 Resource Limits](https://cheatsheetseries.owasp.org/cheatsheets/GraphQL_Cheat_Sheet.html)\n- [OWASP API4:2023 \u2014 Unrestricted Resource Consumption](https://owasp.org/API-Security/editions/2023/en/0xa4-unrestricted-resource-consumption/)\n- [CWE-770: Allocation of Resources Without Limits or Throttling](https://cwe.mitre.org/data/definitions/770.html)\n- [CWE-400: Uncontrolled Resource Consumption](https://cwe.mitre.org/data/definitions/400.html)\n\n## Classification\n\n- **CWE-770** \u2014 Allocation of Resources Without Limits or Throttling\n- **CWE-400** \u2014 Uncontrolled Resource Consumption\n- **OWASP Top 10 2021:** A05:2021 \u2014 Security Misconfiguration\n- **OWASP API Security Top 10:** API4:2023 \u2014 Unrestricted Resource Consumption\n- **STRIDE:** Denial of Service (D)\n\n## Internal Reference\n\nHASI2026141-32 \u2014 Due: 2026-04-16",
"id": "GHSA-h9mw-h4qc-f5jf",
"modified": "2026-04-08T15:05:10Z",
"published": "2026-04-08T15:05:10Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/platform-mesh/kubernetes-graphql-gateway/security/advisories/GHSA-h9mw-h4qc-f5jf"
},
{
"type": "WEB",
"url": "https://github.com/platform-mesh/kubernetes-graphql-gateway/commit/61509656fbab2dbf158f634d6700478ee94221ab"
},
{
"type": "PACKAGE",
"url": "https://github.com/platform-mesh/kubernetes-graphql-gateway"
},
{
"type": "WEB",
"url": "https://github.com/platform-mesh/kubernetes-graphql-gateway/releases/tag/v1.2.9"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
],
"summary": "kubernetes-graphql-gateway: GraphQL Endpoint Vulnerable to Authenticated Denial-of-Service via Unrestricted Query Execution"
}
GHSA-HCC4-C3V8-RX92
Vulnerability from github – Published: 2026-04-01 21:19 – Updated: 2026-04-06 16:46Summary
An unbounded DNS cache could result in excessive memory usage possibly resulting in a DoS situation.
Impact
If an application makes requests to a very large number of hosts, this could cause the DNS cache to continue growing and slowly use excessive amounts of memory.
Patch: https://github.com/aio-libs/aiohttp/commit/c4d77c3533122be353b8afca8e8675e3b4cbda98
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 3.13.3"
},
"package": {
"ecosystem": "PyPI",
"name": "aiohttp"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "3.13.4"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-34513"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": true,
"github_reviewed_at": "2026-04-01T21:19:22Z",
"nvd_published_at": "2026-04-01T21:16:59Z",
"severity": "LOW"
},
"details": "### Summary\n\nAn unbounded DNS cache could result in excessive memory usage possibly resulting in a DoS situation.\n\n### Impact\n\nIf an application makes requests to a very large number of hosts, this could cause the DNS cache to continue growing and slowly use excessive amounts of memory.\n\n-----\n\nPatch: https://github.com/aio-libs/aiohttp/commit/c4d77c3533122be353b8afca8e8675e3b4cbda98",
"id": "GHSA-hcc4-c3v8-rx92",
"modified": "2026-04-06T16:46:44Z",
"published": "2026-04-01T21:19:22Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/aio-libs/aiohttp/security/advisories/GHSA-hcc4-c3v8-rx92"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-34513"
},
{
"type": "WEB",
"url": "https://github.com/aio-libs/aiohttp/commit/c4d77c3533122be353b8afca8e8675e3b4cbda98"
},
{
"type": "PACKAGE",
"url": "https://github.com/aio-libs/aiohttp"
},
{
"type": "WEB",
"url": "https://github.com/aio-libs/aiohttp/releases/tag/v3.13.4"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N/E:U",
"type": "CVSS_V4"
}
],
"summary": "AIOHTTP Affected by Denial of Service (DoS) via Unbounded DNS Cache in TCPConnector"
}
GHSA-HCG3-Q754-CR77
Vulnerability from github – Published: 2025-04-12 00:30 – Updated: 2025-04-14 15:38SSH servers which implement file transfer protocols are vulnerable to a denial of service attack from clients which complete the key exchange slowly, or not at all, causing pending content to be read into memory, but never transmitted.
{
"affected": [
{
"package": {
"ecosystem": "Go",
"name": "golang.org/x/crypto"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.35.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2025-22869"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": true,
"github_reviewed_at": "2025-04-14T15:38:58Z",
"nvd_published_at": "2025-02-26T08:14:24Z",
"severity": "HIGH"
},
"details": "SSH servers which implement file transfer protocols are vulnerable to a denial of service attack from clients which complete the key exchange slowly, or not at all, causing pending content to be read into memory, but never transmitted.",
"id": "GHSA-hcg3-q754-cr77",
"modified": "2025-04-14T15:38:58Z",
"published": "2025-04-12T00:30:26Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2025-22869"
},
{
"type": "WEB",
"url": "https://github.com/golang/crypto/commit/7292932d45d55c7199324ab0027cc86e8198aa22"
},
{
"type": "PACKAGE",
"url": "https://github.com/golang/crypto"
},
{
"type": "WEB",
"url": "https://go-review.googlesource.com/c/crypto/+/652135"
},
{
"type": "WEB",
"url": "https://go.dev/cl/652135"
},
{
"type": "WEB",
"url": "https://go.dev/issue/71931"
},
{
"type": "WEB",
"url": "https://pkg.go.dev/vuln/GO-2025-3487"
},
{
"type": "WEB",
"url": "https://security.netapp.com/advisory/ntap-20250411-0010"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
],
"summary": "golang.org/x/crypto Vulnerable to Denial of Service (DoS) via Slow or Incomplete Key Exchange"
}
GHSA-HCPW-H46R-3C39
Vulnerability from github – Published: 2022-11-01 19:00 – Updated: 2025-05-06 15:30Xenstore: guests can let run xenstored out of memory T[his CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] Malicious guests can cause xenstored to allocate vast amounts of memory, eventually resulting in a Denial of Service (DoS) of xenstored. There are multiple ways how guests can cause large memory allocations in xenstored: - - by issuing new requests to xenstored without reading the responses, causing the responses to be buffered in memory - - by causing large number of watch events to be generated via setting up multiple xenstore watches and then e.g. deleting many xenstore nodes below the watched path - - by creating as many nodes as allowed with the maximum allowed size and path length in as many transactions as possible - - by accessing many nodes inside a transaction
{
"affected": [],
"aliases": [
"CVE-2022-42311"
],
"database_specific": {
"cwe_ids": [
"CWE-401",
"CWE-770"
],
"github_reviewed": false,
"github_reviewed_at": null,
"nvd_published_at": "2022-11-01T13:15:00Z",
"severity": "HIGH"
},
"details": "Xenstore: guests can let run xenstored out of memory T[his CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] Malicious guests can cause xenstored to allocate vast amounts of memory, eventually resulting in a Denial of Service (DoS) of xenstored. There are multiple ways how guests can cause large memory allocations in xenstored: - - by issuing new requests to xenstored without reading the responses, causing the responses to be buffered in memory - - by causing large number of watch events to be generated via setting up multiple xenstore watches and then e.g. deleting many xenstore nodes below the watched path - - by creating as many nodes as allowed with the maximum allowed size and path length in as many transactions as possible - - by accessing many nodes inside a transaction",
"id": "GHSA-hcpw-h46r-3c39",
"modified": "2025-05-06T15:30:38Z",
"published": "2022-11-01T19:00:31Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2022-42311"
},
{
"type": "WEB",
"url": "https://lists.fedoraproject.org/archives/list/package-announce%40lists.fedoraproject.org/message/YTMITQBGC23MSDHUCAPCVGLMVXIBXQTQ"
},
{
"type": "WEB",
"url": "https://lists.fedoraproject.org/archives/list/package-announce%40lists.fedoraproject.org/message/YZVXG7OOOXCX6VIPEMLFDPIPUTFAYWPE"
},
{
"type": "WEB",
"url": "https://lists.fedoraproject.org/archives/list/package-announce%40lists.fedoraproject.org/message/ZLI2NPNEH7CNJO3VZGQNOI4M4EWLNKPZ"
},
{
"type": "WEB",
"url": "https://lists.fedoraproject.org/archives/list/package-announce@lists.fedoraproject.org/message/YTMITQBGC23MSDHUCAPCVGLMVXIBXQTQ"
},
{
"type": "WEB",
"url": "https://lists.fedoraproject.org/archives/list/package-announce@lists.fedoraproject.org/message/YZVXG7OOOXCX6VIPEMLFDPIPUTFAYWPE"
},
{
"type": "WEB",
"url": "https://lists.fedoraproject.org/archives/list/package-announce@lists.fedoraproject.org/message/ZLI2NPNEH7CNJO3VZGQNOI4M4EWLNKPZ"
},
{
"type": "WEB",
"url": "https://www.debian.org/security/2022/dsa-5272"
},
{
"type": "WEB",
"url": "https://xenbits.xenproject.org/xsa/advisory-326.txt"
},
{
"type": "WEB",
"url": "http://xenbits.xen.org/xsa/advisory-326.html"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
]
}
GHSA-HCWQ-8WJF-3GCR
Vulnerability from github – Published: 2026-09-16 22:12 – Updated: 2026-09-16 22:12Summary
The audio decode-duration guard (max_duration_s, env VLLM_MAX_AUDIO_DECODE_DURATION_S, default 600s) that protects against audio decompression-bomb DoS is wired into only the speech-to-text path (/v1/audio/transcriptions). The chat audio path (/v1/chat/completions, input_audio content parts) calls the same decoder with no limit, so an unauthenticated client can submit a few-KB compressed audio file that expands to multiple GB of float32 PCM at decode time, OOM-killing the worker. This is a distinct sibling of CVE-2026-5497 (video frame-count bomb, VideoMediaIO.load_base64) and GHSA-pq5c-rjhq-qp7p (image) in the same media subsystem.
Verified against main at HEAD d78650c (2026-06-16); applicable to the latest release v0.23.0.
Details
The guard rejects long audio during decode (before allocation), implemented in vllm/multimodal/media/audio.py:
- load_audio_pyav — metadata reject (~82-98) and live sample-count reject (~129-136)
- load_audio_soundfile — frames reject (~165-174)
All are gated on if max_duration_s is not None.
It is passed in exactly one place — the transcription serving layer:
# .../speech_to_text/base/serving.py:~170-174
load_audio(buf, sr=..., max_duration_s=self.max_audio_decode_duration_s)
# self.max_audio_decode_duration_s = envs.VLLM_MAX_AUDIO_DECODE_DURATION_S (default 600)
The chat path never threads it:
# vllm/multimodal/media/audio.py:237-238
def load_bytes(self, data: bytes) -> tuple[npt.NDArray, float]:
return load_audio(BytesIO(data), sr=None) # no max_duration_s -> every guard above is skipped
Unauthenticated reachability chain (chat):
parse_input_audio (chat_utils.py) -> parse_audio -> connector.fetch_audio -> AudioMediaIO._load_data_url -> load_base64 -> load_bytes -> load_audio(..., sr=None). The connector never passes max_duration_s, and inline data: URLs need no HTTP fetch (so VLLM_AUDIO_FETCH_TIMEOUT does not bound them). The OpenAI-compatible server has no auth by default (auth only when --api-key / VLLM_API_KEY is set).
Impact
Unauthenticated remote denial of service (availability) via memory amplification on a default-no-auth endpoint, on any deployment serving an audio-capable model. Same class and impact as the sibling CVE-2026-5497 (video). CWE-770 / CWE-409.
Fix
A fix was introduced in this MR: https://github.com/vllm-project/vllm/pull/45908
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 0.23.0"
},
"package": {
"ecosystem": "PyPI",
"name": "vllm"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.24.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-57173"
],
"database_specific": {
"cwe_ids": [
"CWE-770"
],
"github_reviewed": true,
"github_reviewed_at": "2026-09-16T22:12:17Z",
"nvd_published_at": "2026-09-16T17:17:24Z",
"severity": "MODERATE"
},
"details": "### Summary\nThe audio decode-duration guard (`max_duration_s`, env `VLLM_MAX_AUDIO_DECODE_DURATION_S`, default 600s) that protects against audio decompression-bomb DoS is wired into **only** the speech-to-text path (`/v1/audio/transcriptions`). The **chat** audio path (`/v1/chat/completions`, `input_audio` content parts) calls the same decoder with **no** limit, so an **unauthenticated** client can submit a few-KB compressed audio file that expands to multiple GB of float32 PCM at decode time, OOM-killing the worker. This is a distinct sibling of **CVE-2026-5497** (video frame-count bomb, `VideoMediaIO.load_base64`) and **GHSA-pq5c-rjhq-qp7p** (image) in the same media subsystem.\n\nVerified against `main` at HEAD `d78650c` (2026-06-16); applicable to the latest release v0.23.0.\n\n### Details\nThe guard rejects long audio *during* decode (before allocation), implemented in `vllm/multimodal/media/audio.py`:\n- `load_audio_pyav` \u2014 metadata reject (~82-98) and live sample-count reject (~129-136)\n- `load_audio_soundfile` \u2014 frames reject (~165-174)\n\nAll are gated on `if max_duration_s is not None`.\n\nIt is passed in exactly **one** place \u2014 the transcription serving layer:\n```python\n# .../speech_to_text/base/serving.py:~170-174\nload_audio(buf, sr=..., max_duration_s=self.max_audio_decode_duration_s)\n# self.max_audio_decode_duration_s = envs.VLLM_MAX_AUDIO_DECODE_DURATION_S (default 600)\n```\n\nThe chat path never threads it:\n```python\n# vllm/multimodal/media/audio.py:237-238\ndef load_bytes(self, data: bytes) -\u003e tuple[npt.NDArray, float]:\n return load_audio(BytesIO(data), sr=None) # no max_duration_s -\u003e every guard above is skipped\n```\n\nUnauthenticated reachability chain (chat):\n`parse_input_audio` (`chat_utils.py`) -\u003e `parse_audio` -\u003e `connector.fetch_audio` -\u003e `AudioMediaIO._load_data_url` -\u003e `load_base64` -\u003e `load_bytes` -\u003e `load_audio(..., sr=None)`. The connector never passes `max_duration_s`, and inline `data:` URLs need no HTTP fetch (so `VLLM_AUDIO_FETCH_TIMEOUT` does not bound them). The OpenAI-compatible server has no auth by default (auth only when `--api-key` / `VLLM_API_KEY` is set).\n\n### Impact\nUnauthenticated remote denial of service (availability) via memory amplification on a default-no-auth endpoint, on any deployment serving an audio-capable model. Same class and impact as the sibling CVE-2026-5497 (video). CWE-770 / CWE-409.\n\n### Fix\nA fix was introduced in this MR: https://github.com/vllm-project/vllm/pull/45908",
"id": "GHSA-hcwq-8wjf-3gcr",
"modified": "2026-09-16T22:12:18Z",
"published": "2026-09-16T22:12:17Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/vllm-project/vllm/security/advisories/GHSA-hcwq-8wjf-3gcr"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-57173"
},
{
"type": "WEB",
"url": "https://github.com/vllm-project/vllm/pull/45908"
},
{
"type": "WEB",
"url": "https://github.com/vllm-project/vllm/commit/3d20275bb4d434f53055c3c0b645fd8bb072965e"
},
{
"type": "PACKAGE",
"url": "https://github.com/vllm-project/vllm"
},
{
"type": "WEB",
"url": "https://github.com/vllm-project/vllm/releases/tag/v0.24.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
],
"summary": "vLLM: Unauthenticated audio decompression-bomb DoS in /v1/chat/completions"
}
GHSA-HF25-HWQC-V6G2
Vulnerability from github – Published: 2023-05-16 00:30 – Updated: 2024-04-04 04:11In pushDynamicShortcut of ShortcutPackage.java, there is a possible way to get the device into a boot loop due to resource exhaustion. This could lead to local denial of service with no additional execution privileges needed. User interaction is not needed for exploitation.Product: AndroidVersions: Android-11 Android-12 Android-12L Android-13Android ID: A-250576066
{
"affected": [],
"aliases": [
"CVE-2023-20930"
],
"database_specific": {
"cwe_ids": [
"CWE-400",
"CWE-770"
],
"github_reviewed": false,
"github_reviewed_at": null,
"nvd_published_at": "2023-05-15T22:15:11Z",
"severity": "MODERATE"
},
"details": "In pushDynamicShortcut of ShortcutPackage.java, there is a possible way to get the device into a boot loop due to resource exhaustion. This could lead to local denial of service with no additional execution privileges needed. User interaction is not needed for exploitation.Product: AndroidVersions: Android-11 Android-12 Android-12L Android-13Android ID: A-250576066",
"id": "GHSA-hf25-hwqc-v6g2",
"modified": "2024-04-04T04:11:24Z",
"published": "2023-05-16T00:30:17Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2023-20930"
},
{
"type": "WEB",
"url": "https://source.android.com/security/bulletin/2023-05-01"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
]
}
Mitigation
Clearly specify the minimum and maximum expectations for capabilities, and dictate which behaviors are acceptable when resource allocation reaches limits.
Mitigation
Limit the amount of resources that are accessible to unprivileged users. Set per-user limits for resources. Allow the system administrator to define these limits. Be careful to avoid CWE-410.
Mitigation
Design throttling mechanisms into the system architecture. The best protection is to limit the amount of resources that an unauthorized user can cause to be expended. A strong authentication and access control model will help prevent such attacks from occurring in the first place, and it will help the administrator to identify who is committing the abuse. The login application should be protected against DoS attacks as much as possible. Limiting the database access, perhaps by caching result sets, can help minimize the resources expended. To further limit the potential for a DoS attack, consider tracking the rate of requests received from users and blocking requests that exceed a defined rate threshold.
Mitigation MIT-5
Strategy: Input Validation
- Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does.
- When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue."
- Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylists can be useful for detecting potential attacks or determining which inputs are so malformed that they should be rejected outright.
Mitigation MIT-15
For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.
Mitigation
- Mitigation of resource exhaustion attacks requires that the target system either:
- The first of these solutions is an issue in itself though, since it may allow attackers to prevent the use of the system by a particular valid user. If the attacker impersonates the valid user, they may be able to prevent the user from accessing the server in question.
- The second solution can be difficult to effectively institute -- and even when properly done, it does not provide a full solution. It simply requires more resources on the part of the attacker.
- recognizes the attack and denies that user further access for a given amount of time, typically by using increasing time delays
- uniformly throttles all requests in order to make it more difficult to consume resources more quickly than they can again be freed.
Mitigation
Ensure that protocols have specific limits of scale placed on them.
Mitigation MIT-38.1
- If the program must fail, ensure that it fails gracefully (fails closed). There may be a temptation to simply let the program fail poorly in cases such as low memory conditions, but an attacker may be able to assert control before the software has fully exited. Alternately, an uncontrolled failure could cause cascading problems with other downstream components; for example, the program could send a signal to a downstream process so the process immediately knows that a problem has occurred and has a better chance of recovery.
- Ensure that all failures in resource allocation place the system into a safe posture.
Mitigation MIT-47
Strategy: Resource Limitation
- Use quotas or other resource-limiting settings provided by the operating system or environment. For example, when managing system resources in POSIX, setrlimit() can be used to set limits for certain types of resources, and getrlimit() can determine how many resources are available. However, these functions are not available on all operating systems.
- When the current levels get close to the maximum that is defined for the application (see CWE-770), then limit the allocation of further resources to privileged users; alternately, begin releasing resources for less-privileged users. While this mitigation may protect the system from attack, it will not necessarily stop attackers from adversely impacting other users.
- Ensure that the application performs the appropriate error checks and error handling in case resources become unavailable (CWE-703).
CAPEC-125: Flooding
An adversary consumes the resources of a target by rapidly engaging in a large number of interactions with the target. This type of attack generally exposes a weakness in rate limiting or flow. When successful this attack prevents legitimate users from accessing the service and can cause the target to crash. This attack differs from resource depletion through leaks or allocations in that the latter attacks do not rely on the volume of requests made to the target but instead focus on manipulation of the target's operations. The key factor in a flooding attack is the number of requests the adversary can make in a given period of time. The greater this number, the more likely an attack is to succeed against a given target.
CAPEC-130: Excessive Allocation
An adversary causes the target to allocate excessive resources to servicing the attackers' request, thereby reducing the resources available for legitimate services and degrading or denying services. Usually, this attack focuses on memory allocation, but any finite resource on the target could be the attacked, including bandwidth, processing cycles, or other resources. This attack does not attempt to force this allocation through a large number of requests (that would be Resource Depletion through Flooding) but instead uses one or a small number of requests that are carefully formatted to force the target to allocate excessive resources to service this request(s). Often this attack takes advantage of a bug in the target to cause the target to allocate resources vastly beyond what would be needed for a normal request.
CAPEC-147: XML Ping of the Death
An attacker initiates a resource depletion attack where a large number of small XML messages are delivered at a sufficiently rapid rate to cause a denial of service or crash of the target. Transactions such as repetitive SOAP transactions can deplete resources faster than a simple flooding attack because of the additional resources used by the SOAP protocol and the resources necessary to process SOAP messages. The transactions used are immaterial as long as they cause resource utilization on the target. In other words, this is a normal flooding attack augmented by using messages that will require extra processing on the target.
CAPEC-197: Exponential Data Expansion
An adversary submits data to a target application which contains nested exponential data expansion to produce excessively large output. Many data format languages allow the definition of macro-like structures that can be used to simplify the creation of complex structures. However, this capability can be abused to create excessive demands on a processor's CPU and memory. A small number of nested expansions can result in an exponential growth in demands on memory.
CAPEC-229: Serialized Data Parameter Blowup
This attack exploits certain serialized data parsers (e.g., XML, YAML, etc.) which manage data in an inefficient manner. The attacker crafts an serialized data file with multiple configuration parameters in the same dataset. In a vulnerable parser, this results in a denial of service condition where CPU resources are exhausted because of the parsing algorithm. The weakness being exploited is tied to parser implementation and not language specific.
CAPEC-230: Serialized Data with Nested Payloads
Applications often need to transform data in and out of a data format (e.g., XML and YAML) by using a parser. It may be possible for an adversary to inject data that may have an adverse effect on the parser when it is being processed. Many data format languages allow the definition of macro-like structures that can be used to simplify the creation of complex structures. By nesting these structures, causing the data to be repeatedly substituted, an adversary can cause the parser to consume more resources while processing, causing excessive memory consumption and CPU utilization.
CAPEC-231: Oversized Serialized Data Payloads
An adversary injects oversized serialized data payloads into a parser during data processing to produce adverse effects upon the parser such as exhausting system resources and arbitrary code execution.
CAPEC-469: HTTP DoS
An attacker performs flooding at the HTTP level to bring down only a particular web application rather than anything listening on a TCP/IP connection. This denial of service attack requires substantially fewer packets to be sent which makes DoS harder to detect. This is an equivalent of SYN flood in HTTP. The idea is to keep the HTTP session alive indefinitely and then repeat that hundreds of times. This attack targets resource depletion weaknesses in web server software. The web server will wait to attacker's responses on the initiated HTTP sessions while the connection threads are being exhausted.
CAPEC-482: TCP Flood
An adversary may execute a flooding attack using the TCP protocol with the intent to deny legitimate users access to a service. These attacks exploit the weakness within the TCP protocol where there is some state information for the connection the server needs to maintain. This often involves the use of TCP SYN messages.
CAPEC-486: UDP Flood
An adversary may execute a flooding attack using the UDP protocol with the intent to deny legitimate users access to a service by consuming the available network bandwidth. Additionally, firewalls often open a port for each UDP connection destined for a service with an open UDP port, meaning the firewalls in essence save the connection state thus the high packet nature of a UDP flood can also overwhelm resources allocated to the firewall. UDP attacks can also target services like DNS or VoIP which utilize these protocols. Additionally, due to the session-less nature of the UDP protocol, the source of a packet is easily spoofed making it difficult to find the source of the attack.
CAPEC-487: ICMP Flood
An adversary may execute a flooding attack using the ICMP protocol with the intent to deny legitimate users access to a service by consuming the available network bandwidth. A typical attack involves a victim server receiving ICMP packets at a high rate from a wide range of source addresses. Additionally, due to the session-less nature of the ICMP protocol, the source of a packet is easily spoofed making it difficult to find the source of the attack.
CAPEC-488: HTTP Flood
An adversary may execute a flooding attack using the HTTP protocol with the intent to deny legitimate users access to a service by consuming resources at the application layer such as web services and their infrastructure. These attacks use legitimate session-based HTTP GET requests designed to consume large amounts of a server's resources. Since these are legitimate sessions this attack is very difficult to detect.
CAPEC-489: SSL Flood
An adversary may execute a flooding attack using the SSL protocol with the intent to deny legitimate users access to a service by consuming all the available resources on the server side. These attacks take advantage of the asymmetric relationship between the processing power used by the client and the processing power used by the server to create a secure connection. In this manner the attacker can make a large number of HTTPS requests on a low provisioned machine to tie up a disproportionately large number of resources on the server. The clients then continue to keep renegotiating the SSL connection. When multiplied by a large number of attacking machines, this attack can result in a crash or loss of service to legitimate users.
CAPEC-490: Amplification
An adversary may execute an amplification where the size of a response is far greater than that of the request that generates it. The goal of this attack is to use a relatively few resources to create a large amount of traffic against a target server. To execute this attack, an adversary send a request to a 3rd party service, spoofing the source address to be that of the target server. The larger response that is generated by the 3rd party service is then sent to the target server. By sending a large number of initial requests, the adversary can generate a tremendous amount of traffic directed at the target. The greater the discrepancy in size between the initial request and the final payload delivered to the target increased the effectiveness of this attack.
CAPEC-491: Quadratic Data Expansion
An adversary exploits macro-like substitution to cause a denial of service situation due to excessive memory being allocated to fully expand the data. The result of this denial of service could cause the application to freeze or crash. This involves defining a very large entity and using it multiple times in a single entity substitution. CAPEC-197 is a similar attack pattern, but it is easier to discover and defend against. This attack pattern does not perform multi-level substitution and therefore does not obviously appear to consume extensive resources.
CAPEC-493: SOAP Array Blowup
An adversary may execute an attack on a web service that uses SOAP messages in communication. By sending a very large SOAP array declaration to the web service, the attacker forces the web service to allocate space for the array elements before they are parsed by the XML parser. The attacker message is typically small in size containing a large array declaration of say 1,000,000 elements and a couple of array elements. This attack targets exhaustion of the memory resources of the web service.
CAPEC-494: TCP Fragmentation
An adversary may execute a TCP Fragmentation attack against a target with the intention of avoiding filtering rules of network controls, by attempting to fragment the TCP packet such that the headers flag field is pushed into the second fragment which typically is not filtered.
CAPEC-495: UDP Fragmentation
An attacker may execute a UDP Fragmentation attack against a target server in an attempt to consume resources such as bandwidth and CPU. IP fragmentation occurs when an IP datagram is larger than the MTU of the route the datagram has to traverse. Typically the attacker will use large UDP packets over 1500 bytes of data which forces fragmentation as ethernet MTU is 1500 bytes. This attack is a variation on a typical UDP flood but it enables more network bandwidth to be consumed with fewer packets. Additionally it has the potential to consume server CPU resources and fill memory buffers associated with the processing and reassembling of fragmented packets.
CAPEC-496: ICMP Fragmentation
An attacker may execute a ICMP Fragmentation attack against a target with the intention of consuming resources or causing a crash. The attacker crafts a large number of identical fragmented IP packets containing a portion of a fragmented ICMP message. The attacker these sends these messages to a target host which causes the host to become non-responsive. Another vector may be sending a fragmented ICMP message to a target host with incorrect sizes in the header which causes the host to hang.
CAPEC-528: XML Flood
An adversary may execute a flooding attack using XML messages with the intent to deny legitimate users access to a web service. These attacks are accomplished by sending a large number of XML based requests and letting the service attempt to parse each one. In many cases this type of an attack will result in a XML Denial of Service (XDoS) due to an application becoming unstable, freezing, or crashing.