VeloCloud Orchestrator (VCO) on-prem has a security issue where this issue may allow a remote attacker to access privileged internal functionality and impact the VCO host. Successful exploitation may compromise the confidentiality, integrity, and availability of the orchestrator and data managed by the orchestrator.
Hosted, including Dedicated, versions of VCO were impacted and have already been patched.
vLLM through 0.29.0 contains a denial of service vulnerability in the NIXL connector's prefix caching implementation that fails to properly validate block counts across multi-prompt completion requests in prefill/decode disaggregated deployments. Attackers can trigger an assertion failure in NixlBaseConnectorWorker._apply_prefix_caching by submitting completion requests with multiple prompts of varying lengths, causing the decode worker to terminate and become unavailable until restarted.
vLLM through 0.29.0 contains a denial of service vulnerability in P2P KV offloading when OffloadingConnector is configured with TieringOffloadingSpec and a peer-to-peer secondary tier. Attackers can supply arbitrary remote host and port values in kv_transfer_params to create unreachable peer sessions that retain ZeroMQ sockets until the context quota is exhausted, causing an uncaught ZMQError that crashes EngineCore and stops all inference.
vLLM through 0.29.0 contains a resource exhaustion vulnerability in MooncakeConnector where rejected prefill requests create ownerless transfer placeholders that are never reclaimed. Attackers can send rejected requests to exhaust sender task pools, causing valid requests to be delayed by up to 480 seconds while health checks continue returning success.
vLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory. Attackers can supply arbitrary tp_size values in prefill/decode disaggregated deployments to exhaust memory and trigger kernel OOM-kill of the decode worker process.
vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share a single transfer ID in prefill/decode disaggregated deployments. Attackers can trigger GPU memory exhaustion by submitting completion requests with multiple prompts, causing orphaned KV cache blocks to accumulate until process restart and eventually preventing legitimate requests from executing.
vLLM versions through 0.29.0 contain a denial of service vulnerability in the NIXL connector's metadata handling for prefill/decode disaggregated deployments. Attackers can send requests with incomplete kv_transfer_params dictionary entries to trigger an uncaught KeyError in EngineCore scheduling, causing the decode engine to terminate and making all routed requests fail until manual restart.
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.
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.
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 see. 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.