Apache Airflow Vulnerabilities &amp; Security Advisories

High [CVE-2026-68982] Apache Airflow's CeleryExecutor ran callback workloads (`ExecuteCallback`, used for deadline-alert callbacks) without validating the workload's authentication token

Apache Airflow's CeleryExecutor ran callback workloads (`ExecuteCallback`, used for deadline-alert callbacks) without validating the workload's authentication token. Anyone able to write to the Celery broker could enqueue a forged callback and execute arbitrary code on every CeleryExecutor worker, without Airflow credentials. Deployments using the CeleryExecutor are affected. Users are advised to upgrade to apache-airflow 3.4.0 or later, in which the worker claims the callback through the Execution API, validating the token, before running it.

High [CVE-2026-76185] Apache Airflow's Task SDK secret cache built its keys from the team name and secret id with unescaped `_` delimiters, so different (team, secret id) pairs could collide

Apache Airflow's Task SDK secret cache built its keys from the team name and secret id with unescaped `_` delimiters, so different (team, secret id) pairs could collide. In multi-team deployments a Dag author could read another team's Variable or Connection value by crafting a colliding secret id. Only deployments with both `core.multi_team = True` and `secrets.use_cache = True` are affected. Users are advised to upgrade to apache-airflow 3.4.0 or later. Deployments that cannot upgrade can set `secrets.use_cache = False`.

High [CVE-2026-86829] Apache Airflow's scheduler fully deserializes XCom values written by task code while evaluating task dependencies

Apache Airflow's scheduler fully deserializes XCom values written by task code while evaluating task dependencies. A Dag author, who controls what a task writes to XCom, therefore controls data that is deserialized inside the scheduler process - a context the Airflow security model reserves for code installed by the Deployment Manager, above the worker-scoped execution capability a Dag author is documented to have. Deployments that treat Dag authors as less trusted than Deployment Managers are affected, and the scheduler must be evaluating a task dependency that reads a task-written XCom value. Users of apache-airflow are recommended to upgrade to version 3.4.0 or later, which restricts dependency evaluation to primitive XCom values.

High [CVE-2026-107474] Apache Airflow: Connections API: apprise connection "config" extra returned in cleartext

The Apache Airflow connections API did not mask the apprise provider's `config` connection extra, which can embed notification-service credentials. A user with permission to read connections could retrieve it in cleartext, although Airflow 3 treats connection credentials as write-only. Users are advised to upgrade to apache-airflow 3.4.0 or later. Airflow 2.x is also affected; Airflow 2 has reached end of life and will not receive a fix, so Airflow 2 users should upgrade to Apache Airflow 3.4.0 or later.

High8.8ApacheUpdated

High [CVE-2026-111421] Apache Airflow's Task Execution API stored XCom values written by a task without the reserved-key check applied elsewhere, so a Dag author's task could write an XCom that, when a user later viewed it through the API with deserialization enabled, made the API server instantiate classes from the `airflow.*` namespace with arguments the Dag author chose

Apache Airflow's Task Execution API stored XCom values written by a task without the reserved-key check applied elsewhere, so a Dag author's task could write an XCom that, when a user later viewed it through the API with deserialization enabled, made the API server instantiate classes from the `airflow.*` namespace with arguments the Dag author chose. This is an incomplete fix for CVE-2026-59242: users who applied that fix in apache-airflow 3.3.1 remain affected through this route and should upgrade to apache-airflow 3.4.0 or later.

Medium [CVE-2026-50204] Apache Airflow's audit log stored Connection and Variable secret values in cleartext in the `Log.extra` column because the secrets masker matched only on configured sensitive key names

Apache Airflow's audit log stored Connection and Variable secret values in cleartext in the `Log.extra` column because the secrets masker matched only on configured sensitive key names. An authenticated low-privilege user with read access to the event-log / audit API — with no prior access to the affected connection or variable — could read those rows and recover the credentials verbatim. Users should upgrade to apache-airflow 3.4.0 or later, which masks these values and restricts the affected audit rows to authorized users. Airflow 2.x is also affected; Airflow 2 has reached end of life and will not receive a fix, so Airflow 2 users should upgrade to Apache Airflow 3.4.0 or later.

Medium [CVE-2026-82337] Apache Airflow rebuilt deadline alert fields through generic deserialization, which imports and instantiates the class named in the serialized Dag, in the scheduler and API server

Apache Airflow rebuilt deadline alert fields through generic deserialization, which imports and instantiates the class named in the serialized Dag, in the scheduler and API server. A Dag author could cause classes within Airflow's deserialization allow list to be constructed in those processes with arguments they control. Deployments whose Dags use deadline alerts are affected. This is an incomplete fix of CVE-2026-45360, which only hardened the `reference` field; deployments that applied that fix remain affected through the `callback` and `interval` fields. Users of apache-airflow are recommended to upgrade to 3.4.0 or later.

Medium [CVE-2026-96285] Apache Airflow 3.3.1 shipped a fix for CVE-2026-67260 that checks the scheduler's decoded `next_kwargs` value is a dictionary, but performs that check only after deserializing the whole stored payload

Apache Airflow 3.3.1 shipped a fix for CVE-2026-67260 that checks the scheduler's decoded `next_kwargs` value is a dictionary, but performs that check only after deserializing the whole stored payload. Deserialization reconstructs nested objects of classes permitted by the deserialization allow-list — which by default includes Airflow's own classes — and runs their constructors inside the scheduler process before the check can reject the value. A Dag author, who controls a task's `next_kwargs` through the task execution API, can therefore still trigger constructor execution in the scheduler when the `awaiting_input` timeout sweep processes the task, and in the API server when the task instance is served to a worker using an older Execution API version; no non-default configuration is required. Versions before 3.3.0 are not affected. This is an incomplete fix for CVE-2026-67260: deployments that upgraded to 3.3.1 for that issue must also upgrade to apache-airflow 3.4.0 or later.

Medium [CVE-2026-113504] Apache Airflow's secrets masker stopped descending into nested structures after five levels, so a secret registered with the masker — for example a Connection password or a Variable value — that appeared more than five containers deep under non-sensitive keys was shown in cleartext in task logs and rendered template fields

Apache Airflow's secrets masker stopped descending into nested structures after five levels, so a secret registered with the masker — for example a Connection password or a Variable value — that appeared more than five containers deep under non-sensitive keys was shown in cleartext in task logs and rendered template fields. A user who can read task logs or rendered templates, but is not meant to see those secrets, could read them. Users are advised to upgrade to apache-airflow 3.4.0 or later. Airflow 2.x is also affected; Airflow 2 has reached end of life and will not receive a fix, so Airflow 2 users should upgrade to Apache Airflow 3.4.0 or later.

Medium [CVE-2026-111395] Apache Airflow did not reject reserved serialization keys inside a Dag run's `conf`

Apache Airflow did not reject reserved serialization keys inside a Dag run's `conf`. A user allowed to trigger a Dag that uses a deadline alert with an asynchronous callback could pass a `conf` that, once the deadline was missed, either crashed the scheduler repeatedly or was rebuilt into objects inside the Triggerer. Users are advised to upgrade to apache-airflow 3.4.0 or later.

Medium4.3ApacheUpdated

Medium [CVE-2026-111453] Apache Airflow's API endpoint for deleting an asset's queued events checked Dag edit permission against the `dag_id` query parameter, but deleted the queued events of every Dag the caller could read

Apache Airflow's API endpoint for deleting an asset's queued events checked Dag edit permission against the `dag_id` query parameter, but deleted the queued events of every Dag the caller could read. An authenticated user with edit permission on one Dag and read access to others could suppress asset-triggered runs of Dags they cannot edit. This is an incomplete fix for CVE-2026-75157: users who upgraded to apache-airflow 3.3.2 for that fix are still affected and should upgrade to apache-airflow 3.4.0 or later.

Low [CVE-2026-40913] Apache Airflow's public Import Errors API returns the raw parse-failure stack trace to any user with the `ImportError.can_read` permission when the failing file has no registered DAG

Apache Airflow's public Import Errors API returns the raw parse-failure stack trace to any user with the `ImportError.can_read` permission when the failing file has no registered DAG. This happens when a file fails to parse before any DAG is defined, or when all of its previously registered DAGs are removed but the parse error remains. Users who should not see that file's details can read stack traces that may expose file paths and code. Users are advised to upgrade to Apache Airflow 3.4.0 or later, which restricts these stack traces to users with the new `IMPORT_ERRORS_ALL` permission.

Low [CVE-2026-86468] Apache Airflow's CLI records its arguments in the audit log but masks secrets by argument position, so some supported command shapes leave secrets in cleartext, for example a trailing `--description` on Variable commands, `--conn-extra`, and the inline `--password` of `airflow keycloak-auth-manager` commands

Apache Airflow's CLI records its arguments in the audit log but masks secrets by argument position, so some supported command shapes leave secrets in cleartext, for example a trailing `--description` on Variable commands, `--conn-extra`, and the inline `--password` of `airflow keycloak-auth-manager` commands. A user who can read audit events but is denied Variable and Connection access can retrieve those values through the event-log API. Users of apache-airflow are recommended to upgrade to 3.4.0 or later. Audit-log entries written before the upgrade keep the cleartext values: purge the affected entries and rotate the secrets they contain. Airflow 2.x is also affected; Airflow 2 has reached end of life and will not receive a fix, so Airflow 2 users should upgrade to Apache Airflow 3.4.0 or later.

Low [CVE-2026-113508] Apache Airflow's logout endpoint attempted to revoke the presented token on the server, but any failure during revocation was silently ignored and logout still reported success

Apache Airflow's logout endpoint attempted to revoke the presented token on the server, but any failure during revocation was silently ignored and logout still reported success. When the auth manager redirects logout to an external logout URL (for example the FAB or Keycloak auth managers), server-side revocation is the only thing that invalidates the Airflow token; it failed on database errors and, reliably, for tokens from an external issuer accepted via `[api_auth] trusted_jwks_url` whose token identifier is longer than 32 characters. Anyone holding a copy of a user's token from before logout could keep using it against the Airflow API until it expired. Users are advised to upgrade to apache-airflow 3.4.0 or later.

Low [CVE-2026-101031] The `example_dag_decorator` example Dag shipped with Apache Airflow validated an IP address fetched over HTTP with Python's `ipaddress` module and then placed it into a shell command

The `example_dag_decorator` example Dag shipped with Apache Airflow validated an IP address fetched over HTTP with Python's `ipaddress` module and then placed it into a shell command. `ipaddress` accepts IPv6 zone identifiers containing arbitrary characters, so when example Dags are loaded (`load_examples` enabled), a user who can trigger that Dag with a custom `url` parameter could point it at a server whose response injects shell commands run on the worker. This is an incomplete fix for CVE-2025-54941; users who upgraded for CVE-2025-54941 should upgrade again. Users are advised to upgrade to apache-airflow 3.4.0 or later, or to disable example Dags (`load_examples = False`), which is recommended for production.

Low [CVE-2026-111452] Apache Airflow's secrets masker matched sensitive key names only in their snake_case form, so values stored under camelCase keys such as `privateKey`, `connectionString`, `webhookUrl` or `accessKey` — for example in a Connection's extra or a Variable — were not redacted in task logs, rendered template fields or API responses

Apache Airflow's secrets masker matched sensitive key names only in their snake_case form, so values stored under camelCase keys such as `privateKey`, `connectionString`, `webhookUrl` or `accessKey` — for example in a Connection's extra or a Variable — were not redacted in task logs, rendered template fields or API responses. A user who can read task logs or rendered templates, but is not meant to see those credentials, could read them in cleartext. Users are advised to upgrade to apache-airflow 3.4.0 or later. Airflow 2.x is also affected; Airflow 2 has reached end of life and will not receive a fix, so Airflow 2 users should upgrade to Apache Airflow 3.4.0 or later.

Low [CVE-2026-111454] Apache Airflow: API: asset endpoints return scheduled/producing/consuming Dag IDs the caller cannot read

Apache Airflow's asset API endpoints returned the IDs of Dags that are scheduled on, produce, or consume an asset without checking whether the caller may read those Dags. An authenticated user with asset read access could learn the IDs of Dags they are not authorized to Users are advised to upgrade to apache-airflow 3.4.0 or later. Airflow 2.x is also affected; Airflow 2 has reached end of life and will not receive a fix, so Airflow 2 users should upgrade to Apache Airflow 3.4.0 or later.

Low [CVE-2026-111455] Apache Airflow's asynchronous connection test checked permissions when the test was queued — only create permission when the connection ID did not exist yet — but, when the test completed with `commit_on_success`, wrote the tested connection under that ID

Apache Airflow's asynchronous connection test checked permissions when the test was queued — only create permission when the connection ID did not exist yet — but, when the test completed with `commit_on_success`, wrote the tested connection under that ID. A user allowed to create but not edit connections could overwrite a connection someone else had created in the meantime. Only deployments that enable the non-default `[core] test_connection` setting are affected. Users are advised to upgrade to apache-airflow 3.4.0 or later.

High [CVE-2026-111422] In apache-airflow-providers-ssh before 7.0.1, SSHRemoteJobOperator quoted the values of its templated `environment` parameter only once, although they are placed inside an already-quoted wrapper script, so shell syntax such as `$(...)` in a value was executed on the remote SSH host

In apache-airflow-providers-ssh before 7.0.1, SSHRemoteJobOperator quoted the values of its templated `environment` parameter only once, although they are placed inside an already-quoted wrapper script, so shell syntax such as `$(...)` in a value was executed on the remote SSH host. In a Dag that passes trigger-time input such as `params` or `dag_run.conf` into `environment`, a user who can trigger that Dag but cannot author Dags could run arbitrary commands on the remote host as the SSH connection's user. Users are advised to upgrade apache-airflow-providers-ssh to 7.0.1 or later. Affected product named by the advisory: Apache Airflow SSH provider.

UnratedApache

Advisory [CVE-2026-86467] Apache Airflow Amazon provider: AWS auth manager: SAML login CSRF logs users into the attacker's account

Apache Airflow Amazon provider: the AWS auth manager's SAML login flow does not bind the authentication response to the browser that began it. The login handler discards the AuthnRequest ID and the callback validates the assertion without requiring a matching InResponseTo, so an unauthenticated attacker who can cause a victim to load a page can complete a SAML login in the victim's browser using the attacker's own assertion, silently logging the victim into the attacker's Airflow account. Subsequent work the victim performs lands in a session the attacker controls. The victim needs only to visit an attacker-controlled page; no Airflow account or credential is required of the attacker. This is the same login-CSRF class fixed in the Keycloak auth manager as CVE-2026-40948. Users of apache-airflow-providers-amazon are recommended to upgrade to version 9.38.0, which fixes the issue.

Critical [CVE-2026-94134 +1] The Apache Airflow SSH provider hook disables SSH host-key verification by default: unless `host_key` is configured or `no_host_key_check` is explicitly set to false, it accepts any server key

The Apache Airflow SSH provider hook disables SSH host-key verification by default: unless `host_key` is configured or `no_host_key_check` is explicitly set to false, it accepts any server key. This affects SSH and SFTP connections that rely on the default. An attacker in a machine-in-the-middle position between the worker and the SSH or SFTP server can impersonate the server and capture the credentials or data a task sends. Users of apache-airflow-providers-ssh and apache-airflow-providers-sftp are recommended to upgrade to apache-airflow-providers-ssh 7.0.0 and apache-airflow-providers-sftp 7.0.0 or later. The deferrable SFTP code paths had the same default and are covered separately by CVE-2026-94135; apache-airflow-providers-sftp 7.0.0 fixes both. Affected products named by the advisory: Apache Airflow SFTP provider.

Medium [CVE-2026-94134 +1] The Apache Airflow SFTP provider's deferrable code paths (`SFTPTrigger`, `SFTPHookAsync`, `SFTPClientPool`) disable SSH host-key verification by default unless the connection explicitly sets `no_host_key_check` to false

The Apache Airflow SFTP provider's deferrable code paths (`SFTPTrigger`, `SFTPHookAsync`, `SFTPClientPool`) disable SSH host-key verification by default unless the connection explicitly sets `no_host_key_check` to false. Deployments that run SFTP sensors or operators in deferrable mode are affected. An attacker in a machine-in-the-middle position between the triggerer and the SFTP server can impersonate the server and capture the credentials or data exchanged. Users of apache-airflow-providers-sftp are recommended to upgrade to apache-airflow-providers-sftp 7.0.0 or later. This is a separate defect from CVE-2026-94134 (the synchronous SSH hook default, which also affects the SFTP hook); users who applied the fix for one also need this release for the other - apache-airflow-providers-sftp 7.0.0 fixes both.

Medium5.8Apache

Medium [CVE-2026-96806] The Apache Airflow Keycloak auth manager did not end the session on logout: the logout route revoked the presented token, but `KeycloakJWTMiddleware` then issued a new, unrevoked token on the same response, so the user stayed logged in

The Apache Airflow Keycloak auth manager did not end the session on logout: the logout route revoked the presented token, but `KeycloakJWTMiddleware` then issued a new, unrevoked token on the same response, so the user stayed logged in. Deployments using the Keycloak auth manager are affected: after a user logs out, anyone with later access to the same browser (for example on a shared workstation) can continue to act as that user. Users of apache-airflow-providers-keycloak are recommended to upgrade to apache-airflow-providers-keycloak 0.11.1 or later. Affected product named by the advisory: Apache Airflow Keycloak provider.

Medium [CVE-2026-92228] The Apache Airflow Keycloak provider's multi-team provisioning (`_attach_team_permissions`) granted every team role global read access to audit logs

The Apache Airflow Keycloak provider's multi-team provisioning (`_attach_team_permissions`) granted every team role global read access to audit logs. Deployments that enable multi-team mode with the Keycloak auth manager and provision teams through it are affected: any authenticated member of one team could read other teams' audit events through the UI or the event-log API. Users of apache-airflow-providers-keycloak are recommended to upgrade to apache-airflow-providers-keycloak 0.11.1 or later.

Medium6.3Apache

Medium [CVE-2026-86843] The Apache Airflow Teradata provider's compute-cluster example Dag declared every one of its Dag Params as unconstrained free text and templated them straight into the compute-cluster operators, which interpolate those values into Teradata DDL

The Apache Airflow Teradata provider's compute-cluster example Dag declared every one of its Dag Params as unconstrained free text and templated them straight into the compute-cluster operators, which interpolate those values into Teradata DDL. A user who is permitted to trigger that Dag - a lower-trust role than the Dag author, and one that needs no Teradata credentials of its own - could therefore supply SQL fragments that execute under the connection the task runs as, and could additionally redirect the task at any other connection defined in the deployment, because the connection id was itself a free-text Param. Only deployments that run this example Dag, or a Dag copied from it, are affected; the provider's operator code is unchanged. Users of apache-airflow-providers-teradata are recommended to upgrade to version 3.7.0 or later, whose example constrains the Params to validated identifiers and a closed value set and removes connection selection and free-form option strings from trigger-time input. Upgrading does not change a Dag already copied from the example; users who copied it should apply the same constraints to their copy.

Medium6.3Apache

Medium [CVE-2026-81930] Apache Airflow's Snowflake provider did not validate the connection's `account` and `region` fields before interpolating them into request URLs

Apache Airflow's Snowflake provider did not validate the connection's `account` and `region` fields before interpolating them into request URLs. The SQL API endpoint is built as `https://{account}.snowflakecomputing.com/api/v2/statements`, so an `account` value containing `/`, `?` or `#` demotes the intended domain to a path, query or fragment and leaves the attacker in control of the request host. The provider sends that request with an `Authorization: Bearer` header carrying a JWT minted from the connection's private key, or the configured OAuth or programmatic access token. A user who can edit the Snowflake connection but cannot read its secrets — Airflow gives connection-configuration users write-only access to stored credentials, and a `private_key_file` lives on the worker rather than in the connection — can therefore cause a valid token for the account to be delivered to a host of their choosing and replay it against the genuine Snowflake endpoint. No Dag-authoring ability is required: the attacker edits the connection and waits for an existing Dag to use it. The same unvalidated value was also used to build the OAuth token-request URL and the Cortex Agent base URL. Affects deployments where Snowflake connections are editable by users who are not trusted with the connection's credentials.

Medium [CVE-2026-81862] Apache Airflow's Teradata provider embedded cloud storage credentials directly into SQL statements

Apache Airflow's Teradata provider embedded cloud storage credentials directly into SQL statements. `S3ToTeradataOperator` and `AzureBlobStorageToTeradataOperator` interpolate the source bucket's credentials as plain string literals into the `CREATE MULTISET TABLE... LOCATION` statement whenever the bucket is private and no `teradata_authorization_name` is configured — which is the default credential path for both operators. The statement is then logged and executed, so the credentials reach two places outside the operator's control. The two operators expose different credentials through different channels, and deployments should check both. `S3ToTeradataOperator` takes its values from `s3_hook.get_credentials()`, which under an instance profile or IRSA returns runtime AWS credentials that were never registered with Airflow's secrets masker — and the STS session token is runtime-generated and therefore unmasked even when an AWS connection is configured. Those credentials appear **in the Airflow task log**, readable by any user with log-view permission on the Dag.

Medium [CVE-2026-81914] Apache Airflow's Google provider built Google Drive search expressions by interpolating file and folder names directly into single-quoted string literals, without escaping the quote character that delimits them

Apache Airflow's Google provider built Google Drive search expressions by interpolating file and folder names directly into single-quoted string literals, without escaping the quote character that delimits them. A name containing an apostrophe therefore terminated the literal early and appended clauses of the attacker's choosing to the query. The names are frequently not written by the Dag author. In a wildcard `gcs_to_gdrive` transfer they come from the source bucket listing, so anyone able to create objects in that bucket controls them — typically an external data producer or an ingest-only service account, a different trust principal from the Dag author. An injected clause can broaden the match and so steer which file or folder the hook resolves: an upload can be directed into a folder the attacker named, and, because downloads select the most recently modified match, a download can return a file they placed rather than the one the Dag asked for. Affects deployments passing externally-sourced names to the Google Drive hook, including wildcard `gcs_to_gdrive` transfers from buckets writable by less-trusted principals. Users are advised to upgrade to `apache-airflow-providers-google` `22.6.0` or later, which escapes quote and backslash characters in every value interpolated into a Drive query.

Medium [CVE-2026-68870 +4] Apache Airflow HashiCorp provider: the HashiCorp Vault secrets backend's team-scope guard can be bypassed with a user-controlled key

Apache Airflow HashiCorp provider: the HashiCorp Vault secrets backend's team-scope guard can be bypassed with a user-controlled key. In a multi-team deployment, a Dag author scoped to one team can supply a Variable key containing a path separator that causes the backend to resolve a secret belonging to a different team, because after the team-scoped lookup misses the backend falls back to a team-agnostic path concatenated from the unvalidated key. The Execution API Variables route accepts a path-shaped key, so this is reachable from ordinary Dag code. Affects multi-team deployments using the HashiCorp Vault secrets backend. Single-team deployments are not affected, as there is no cross-team boundary to cross. This is the same class as CVE-2026-86465, CVE-2026-68870, CVE-2026-68871 and CVE-2026-68872 in the Akeyless, Azure Key Vault, Yandex Lockbox and Amazon secrets backends. Users of apache-airflow-providers-hashicorp are recommended to upgrade to version 4.8.0 or later, which fixes the issue.

Critical [CVE-2026-86473] Apache Airflow: Logout ignores a presented Authorization bearer token, leaving it revocable only by expiry

Apache Airflow: the Core API logout endpoint revokes only a session token presented as the _token cookie. When a client logs out presenting its credential as an Authorization bearer header instead, the endpoint returns its normal logout response but revokes nothing, so the token remains valid until it expires. An attacker who already holds a copy of that token keeps the victim's access after the victim has logged out and believes the session ended; the default token lifetime is 24 hours and is configurable. Affects API clients that authenticate with a bearer token rather than the browser session cookie. The attacker must already possess a copy of a valid token; obtaining one is outside the scope of this issue, and no privileges beyond the victim's own are gained. Users of apache-airflow are recommended to upgrade to apache-airflow version 3.3.2 or later, which fixes the issue.

Medium [CVE-2026-75158] Apache Airflow: Assets events API returns asset events for every Dag with no per-Dag authorization filter

Apache Airflow's `/assets/events` API returned asset events for every Dag in the deployment, with no filter restricting them to the Dags the caller is authorized to read. Any authenticated user holding asset-read access could therefore enumerate asset events — including the source Dag ID, task ID, run ID and event timestamps — for Dags they have no permission to Because the filter was also absent from the count query, `total_entries` and pagination disclosed the existence of hidden Dags even without inspecting individual rows. Deployments are affected whenever per-Dag access control is used to separate teams or tenants; no special configuration is required. Upgrade to apache-airflow 3.3.2 or later.

Medium4.2Apache

Medium [CVE-2026-82355] Apache Airflow: Session cookie silently overrides explicit Authorization bearer header, enabling session fixation

When a request to the Airflow core API carries both a session cookie and an explicit `Authorization: Bearer` token, Airflow resolves the caller from the cookie and ignores the bearer token, inverting the intended precedence of bearer over cookie. The request then executes -- and is recorded in the audit log -- as the cookie's principal rather than the identity the client explicitly presented. Only Apache Airflow 3.3.0 and 3.3.1 are affected. Earlier releases do not contain the code path that caches the cookie-derived user, and are not vulnerable. Exploiting this requires an attacker to first place a valid session cookie of their own into the victim's browser or client: for example by cookie tossing from a sibling subdomain, through cross-site scripting in a separate application sharing a parent domain, or via a shared workstation. Deployments that host the Airflow UI on a domain shared with other applications are therefore the most exposed; a deployment on a dedicated domain with no co-hosted applications is not reachable this way. The consequence is principal confusion and misattributed audit records rather than a direct privilege escalation. Users of 3.3.0 or 3.3.1 should upgrade to Apache Airflow 3.3.2 or later, which resolves the caller from the explicitly supplied credential whenever one is present.

High7.5Apache

High [CVE-2026-75157] Apache Airflow: Asset queued-events DELETE endpoints gated on Dag READ instead of Dag EDIT (asset-triggered scheduling suppression)

Apache Airflow's asset queued-events DELETE endpoints checked the caller's Dag-axis permission with `READ` instead of `EDIT`. Any authenticated user who could read a Dag could therefore delete that Dag's queued asset events, silently suppressing asset-triggered scheduling for it — a state-changing action gated on a read-only permission. Deployments are affected whenever asset-triggered scheduling is in use and Dag read access is granted more widely than Dag edit access, which is the normal RBAC arrangement; no special configuration is required. Upgrade to apache-airflow 3.3.2 or later.

Critical9.8Apache

Critical [CVE-2026-76187] Apache Airflow Keycloak provider: the unauthenticated token endpoint accepts a client-credentials grant for any confidential client registered in the Keycloak realm, not only the client configured for Airflow

Apache Airflow Keycloak provider: the unauthenticated token endpoint accepts a client-credentials grant for any confidential client registered in the Keycloak realm, not only the client configured for Airflow. No allowlist restricts which client ids may authenticate, so the credentials of an unrelated application that happens to share the realm are valid Airflow login credentials, and Airflow mints a signed session token for that application's service account. The endpoint also answers unauthenticated credential guesses against Keycloak under Airflow's identity. Affects deployments using the Keycloak auth manager whose realm is shared with other confidential clients. The attacker needs valid credentials for any one of those clients, not for Airflow. Resource authorization is still evaluated per subject, so the access gained is whatever that service account holds, plus any endpoint gated only on being authenticated. Users of apache-airflow-providers-keycloak are recommended to upgrade to version 0.10.0 or later, which accepts only the configured client on that grant.

Critical [CVE-2026-76186] Apache Airflow Keycloak provider: from Airflow 3.3 the Keycloak auth manager takes a user's identity from the signed Airflow session token but takes the Keycloak access and refresh tokens used for every authorization decision from separate, unauthenticated cookies, and never checks that the two describe the same subject

Apache Airflow Keycloak provider: from Airflow 3.3 the Keycloak auth manager takes a user's identity from the signed Airflow session token but takes the Keycloak access and refresh tokens used for every authorization decision from separate, unauthenticated cookies, and never checks that the two describe the same subject. A user who holds any valid Airflow login of their own, together with another subject's Keycloak access or refresh token obtained out of band, can pair the two: Airflow then authorizes requests with the foreign token's privileges while the session identity, audit log and cache keys continue to name the attacker's own account. The refresh path re-issues an Airflow session token for the original identity carrying the foreign tokens, so the mismatched pairing survives across sessions. Affects deployments running Airflow 3.3 or later with the Keycloak auth manager. Earlier versions carried the Keycloak tokens inside the signed session token, so the binding existed and was lost when they moved into separate cookies. Users of apache-airflow-providers-keycloak are recommended to upgrade to version 0.10.0 or later, which binds the cookie-supplied tokens to the session identity.

Critical [CVE-2026-82311 +1] Apache Airflow FAB provider: changing a user's password through the Admin user-edit PATCH endpoint does not invalidate that user's existing database-backed sessions

Apache Airflow FAB provider: changing a user's password through the Admin user-edit PATCH endpoint does not invalidate that user's existing database-backed sessions. An attacker who already holds a copy of the victim's session cookie keeps full access as that user after the password change, so the password reset does not evict them. Affects deployments using the FAB auth manager with database-backed sessions; an administrator (or the user themselves) performing a routine password change is the trigger, and no attacker interaction with the endpoint is needed. This is a second, independent route to the outcome addressed by CVE-2026-82311, which corrected an identifier comparison in the session-invalidation helper. That fix does not repair this endpoint, because the PATCH path never calls the helper at all. Deployments that applied the CVE-2026-82311 fix must also upgrade for this one. Users of apache-airflow-providers-fab are recommended to upgrade to version 3.9.0 or later, which fixes the issue.

Critical9.8Apache

Critical [CVE-2026-82311 +1] Apache Airflow FAB provider: resetting a user's password does not delete that user's existing database-backed sessions, despite documented behaviour that it does

Apache Airflow FAB provider: resetting a user's password does not delete that user's existing database-backed sessions, despite documented behaviour that it does. The cleanup compares the string identifier Flask-Login stores in the session against the user's integer database identifier, so the comparison never matches and no session is removed. An attacker who already holds a copy of the victim's session cookie keeps access as that user after the password change, so the reset does not evict them. Affects deployments using the FAB auth manager with `[fab] session_backend=database`. The trigger is an administrator (or the user) running the supported password-reset command as a containment action after a session cookie has been compromised; the secure-cookie backend is out of scope, as it documents that it cannot centrally delete sessions. apache-airflow-providers-fab 3.9.0 also fixes CVE-2026-86462, a second, independent route to the same outcome via the Admin user-edit endpoint; a single upgrade closes both. Users of apache-airflow-providers-fab are recommended to upgrade to version 3.9.0 or later, which compares the identifiers consistently.

High8.1Apache

High [CVE-2026-75156 +1] Apache Airflow FAB provider: the Authentik OAuth path in the FAB auth manager does not validate the issuer or audience claims of the id_token it accepts

Apache Airflow FAB provider: the Authentik OAuth path in the FAB auth manager does not validate the issuer or audience claims of the id_token it accepts. An attacker holding a token that the same Authentik identity provider minted for a different client application can present it to Airflow and be authenticated as the user it names, because the audience claim is never checked. Affects deployments using the FAB auth manager with Authentik OAuth where the same Authentik instance also serves other applications; the attacker needs a valid token for any of those other applications, not for Airflow. CVE-2026-75156 corrected the same missing validation on the Azure AD path in this file; the Authentik path was left unchanged and is fixed here. Deployments that applied the CVE-2026-75156 fix and use Authentik must also upgrade for this one. Users of apache-airflow-providers-fab are recommended to upgrade to version 3.9.0 or later, which fixes the issue.

High7.2Apache

High [CVE-2026-82310] Apache Airflow FAB provider: deactivating a user account does not stop tokens issued to that account before deactivation

Apache Airflow FAB provider: deactivating a user account does not stop tokens issued to that account before deactivation. Password authentication correctly rejects the disabled account, but the Core API continues to accept an existing, unexpired token naming it, and lets that token mint a replacement — so the account keeps its role-scoped access indefinitely after an administrator has disabled it. The user replays their own legitimate credential; no signature forgery or privilege escalation is involved, and the access stays within the roles the account already held. Affects deployments using Airflow 3 with the FAB auth manager and Core API token authentication, where an administrator deactivates an account whose row remains in the database and whose previously issued token has not expired. The trigger is administrative deactivation as a containment action, which silently fails to contain. Users of apache-airflow-providers-fab are recommended to upgrade to version 3.9.0 or later, which rejects tokens naming a deactivated account.

High [CVE-2026-86792] Apache Airflow Apache Kafka provider versions 1.15.0 before 2.0.0 resolve dotted-path strings found in a Kafka connection's `extra` field into Python callables via `import_string`, with no allowlist, and hand them to the confluent-kafka client which invokes them

Apache Airflow Apache Kafka provider versions 1.15.0 before 2.0.0 resolve dotted-path strings found in a Kafka connection's `extra` field into Python callables via `import_string`, with no allowlist, and hand them to the confluent-kafka client which invokes them. Deployments that have enabled the Kafka event producer — `dag_run_events_enabled` or `task_instance_events_enabled`, both disabled by default — build that client inside the scheduler process, so a user whose only privilege is editing Airflow connections gains arbitrary code execution in the control plane; the Airflow security model limits connection-configuration users to code execution on workers, not the scheduler. Deployments using Google Managed Kafka are not affected, because that code path overwrites any user-supplied `oauth_cb`; plain brokers and Amazon MSK are exposed. Users are recommended to upgrade to apache-airflow-providers-apache-kafka 2.0.0 or later, which adds an allowlist configuration option for connection-string callbacks.

Medium [CVE-2026-68870 +3] Apache Airflow Akeyless provider: the Akeyless secrets backend's team-scope guard can be bypassed with a user-controlled key

Apache Airflow Akeyless provider: the Akeyless secrets backend's team-scope guard can be bypassed with a user-controlled key. In a multi-team deployment, a Dag author scoped to one team can supply a Variable key containing a path separator that causes the backend to resolve a secret belonging to a different team, because the lookup path is concatenated from an unvalidated key after the team-scoped lookup misses. The Execution API Variables route accepts a path-shaped key, so this is reachable from ordinary Dag code. Affects multi-team deployments using the Akeyless secrets backend. Single-team deployments are not affected, as there is no cross-team boundary to cross. This is the same class as CVE-2026-68870, CVE-2026-68871 and CVE-2026-68872 in the Azure Key Vault, Yandex Lockbox and Amazon secrets backends. Users of apache-airflow-providers-akeyless are recommended to upgrade to version 0.3.1 or later, which fixes the issue.

Critical [CVE-2026-59243 +1] Apache Airflow FAB provider: FAB Azure AD OAuth: id_token issuer/audience not validated — cross-tenant authentication bypass

Apache Airflow FAB provider versions 3.7.3 through 3.8.0 do not validate the issuer or audience of Azure AD `id_token`s during OAuth login. Deployments are affected only when the FAB auth manager is configured with Azure AD as an OAuth provider. Because the signing keys are fetched from Microsoft's **multi-tenant** JWKS endpoint, an `id_token` minted in *any* Azure tenant — including one the attacker creates — passes signature verification, and the username and role assignments are then read from that attacker-controlled token. Anyone able to register an Azure tenant can therefore authenticate to the Airflow UI with no prior access to the deployment. The fix for **CVE-2026-59243** was incomplete, and this advisory closes the remaining gap: that fix made the provider verify the `id_token` signature, but did not add issuer or audience checks. Operators who already applied the CVE-2026-59243 fix are **still affected and must upgrade again** — 3.7.3 is the release that shipped that fix, so every version containing it falls inside this affected range. Upgrade to apache-airflow-providers-fab `3.8.1` or later.

High [CVE-2026-58076] Apache Airflow: Unguarded import_string of airflow_exc_ser / base_exc_ser exception nodes in BaseSerialization.deserialize enables DAG-author RCE on Scheduler / API Server

Apache Airflow's serialization layer reconstructed exception nodes by calling `import_string()` on a class name taken from the serialized blob and instantiating it with arguments from the same blob, with no restriction on what could be imported. An operator's `executor_config` reaches that branch, so a Dag author could place a value there that causes an arbitrary callable to be imported and invoked -- for example `subprocess.check_output`, or `builtins.eval` on the `builtins`-prefixed variant. The code runs in the **Scheduler**, which reconstructs serialized Dags in its normal loop with no request involved, and in the **API server**, on any authenticated read of the Dag such as `GET /api/v2/dags/{dag_id}/details`. Both are components the Airflow security model states must never execute Dag-author code, and both hold the metadata database credentials and the JWT signing secret. No non-default configuration is required. This is a **different sink from CVE-2026-33264**, which covered only the trigger branch of the same deserializer: deployments that upgraded in response to that advisory are still affected through the exception branch and must upgrade again. Users are advised to upgrade to apache-airflow 3.3.1 or later, which restricts the imported class to a subclass of `BaseException`.

High7.3Apache

High [CVE-2026-67260] Apache Airflow: DAG-author remote code execution on the Scheduler via awaiting_input next_kwargs deserialization

Apache Airflow 3.3.0 moved human-in-the-loop tasks from the triggerer to a new `awaiting_input` task state swept by the scheduler. That sweep deserializes the task instance's `next_kwargs` without an allow-list, so a Dag author — who controls that value through the task execution API — can cause an arbitrary module import and object instantiation inside the scheduler process, or terminate the scheduler job. No non-default configuration is required: the sweep runs unconditionally every 15 seconds, and the default `allowed_deserialization_classes` setting does not cover this code path. Versions before 3.3.0 are not affected, because human-in-the-loop tasks deferred onto the triggerer instead. This is a different code path from CVE-2026-58076, which covers the same unguarded exception-node deserialization reached elsewhere — deployments that applied that fix must upgrade for this issue as well. Users are advised to upgrade to apache-airflow 3.3.1 or later.

High [CVE-2026-67587] Apache Airflow: DAG-author remote code execution on the Scheduler via a Serde `Callback` deserialization gadget

Apache Airflow's Task SDK rebuilt a `Callback` object from serialized data by re-running its constructor, which imports the module named by the stored callback path. Because `SyncCallback` is itself an Airflow class it passes the default `allowed_deserialization_classes` allow-list, so tightening that setting does not help. A Dag author — who controls a task instance's `next_kwargs` through the task execution API — can therefore cause an arbitrary module to be imported inside the scheduler process, when the scheduler's `awaiting_input` timeout sweep deserializes that value. No non-default configuration is required; the sweep runs unconditionally. Versions before 3.3.0 are not affected: the class existed, but the scheduler sweep that reaches it did not. This is a separate code path from CVE-2026-58076 and CVE-2026-67260, which cover different gadgets reaching deserialization — applying either of those fixes does not address this one. Users are advised to upgrade to apache-airflow 3.3.1 or later.

High7.5Apache

High [CVE-2026-68968] Apache Airflow: Authorization bypass in the Backfill API through conflicting interpretations of the backfill id

Apache Airflow's Backfill API authorized a request against a Dag id supplied by the caller whenever the `backfill_id` path segment failed to parse. The authorization dependency parsed it with `int()` while the route handler parsed it as pydantic's `NonNegativeInt`, which accepts values `int()` rejects (`1.0` coerces to `1`); FastAPI resolves dependencies before endpoint validation, so the two acted on different Dags. An authenticated user holding edit permission on any single Dag could therefore read, pause and cancel backfills belonging to any other Dag, including moving another Dag's queued runs to `failed`. No non-default configuration is required and backfill ids are sequential, so finding a target is trivial. Users are advised to upgrade to apache-airflow 3.3.1 or later, which parses the backfill id with the same type the routes declare.

Medium [CVE-2026-59244] Apache Airflow: Secrets masker: `var.json` Variable values not masked in the Rendered Templates UI

Apache Airflow's secrets masker did not mask `var.json` Variable values whose value is a dict in the Rendered Templates UI — the dict value failed an `isinstance(str)` guard — so a secret stored as a JSON Variable and referenced in a template via `var.json` was displayed in cleartext to any user with access to that task's Rendered Templates view. Users are advised to upgrade to apache-airflow 3.3.1 or later, which masks nested Variable values regardless of type.

Medium5.4Apache

Medium [CVE-2026-59242] Apache Airflow: Arbitrary airflow.* class instantiation on the API server via the XCom deserialize endpoint

Apache Airflow's XCom `GET /api/v2/{...}/xcomEntries/{key}?deserialize=true` endpoint passed a string-literal payload through `BaseXCom.deserialize_value` without the `_check_forbidden_xcom_keys` guard, allowing an authenticated API user with XCom write-and-read access to instantiate arbitrary `airflow.*` classes on the API server (CWE-502). An authenticated user who can write an XCom value and then read it back with `deserialize=true` triggers the unsafe instantiation. Users are advised to upgrade to apache-airflow 3.3.1 or later, which rejects reserved XCom serialization keys submitted as JSON string literals.

Medium [CVE-2026-54183] Apache Airflow: Airflow Variables were not masked in the UI for authenticated users

Apache Airflow's secrets masker hides values stored under sensitive key names when they are displayed in the UI. The masker's recursion-depth limit did not descend into values nested inside a list, tuple, or set beyond that limit, so an Airflow Variable holding such a deeply-nested value was shown unmasked in the Variables UI. The exposure is limited to the UI: any authenticated user who can see the Variable in the UI can already read its full value through the Variables REST API, so this does not disclose data the user could not otherwise obtain — the masking is a shoulder-surfing defense for the UI, not an access-control boundary. This is an incomplete-fix follow-up to CVE-2026-42358, whose fix made only the dictionary walk unbounded; lists, tuples, and sets beyond the depth limit remained unmasked in the UI. Deployments that applied the CVE-2026-42358 fix should also upgrade to address this residual case. Upgrade to apache-airflow 3.3.1 or later.