-
CVE-2026-94635
•
published on October 2, 2026
Allocation of resources without limits or throttling, Improper handling of length parameter inconsistency vulnerability in Apache Thrift Lua bindings.
This issue affects Apache Thrift: before 0.25.0.
Users are recommended to upgrade to version 0.25.0, which fixes the issue.
-
CVE-2026-94634
•
published on October 2, 2026
Allocation of resources without limits or throttling, Initialization of a resource with an insecure default vulnerability in Apache Thrift Python bindings.
This issue affects Apache Thrift: before 0.25.0.
Users are recommended to upgrade to version 0.25.0, which fixes the issue.
-
CVE-2026-94633
•
published on October 2, 2026
Memory allocation with excessive size value, Improper handling of length parameter inconsistency vulnerability in Apache Thrift Dart bindings.
This issue affects Apache Thrift: before 0.25.0.
Users are recommended to upgrade to version 0.25.0, which fixes the issue.
-
CVE-2026-94632
•
published on October 9, 2026
Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting') vulnerability in Stiofan BlockStrap Page Builder - Bootstrap Blocks blockstrap-page-builder-blocks allows Reflected XSS.This issue affects BlockStrap Page Builder - Bootstrap Blocks: from n/a through 0.1.58.
-
CVE-2026-94627
•
published on September 21, 2026
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.
-
CVE-2026-94626
•
published on September 21, 2026
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.
-
CVE-2026-94625
•
published on September 21, 2026
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.
-
CVE-2026-94624
•
published on September 21, 2026
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.
-
CVE-2026-94623
•
published on September 21, 2026
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.
-
CVE-2026-94622
•
published on September 21, 2026
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.
-
CVE-2026-94621
•
published on October 15, 2026
AliasVault is a privacy-first password manager with built-in email aliasing. Prior to 0.30.3, the AliasVault server API favicon extractor validates a user-supplied hostname with a first DNS lookup but allows the HTTP client to perform a second DNS lookup when connecting. An authenticated low-privileged user operating an attacker-controlled authoritative DNS server can return a public address during validation and a private address during connection, causing a publicly reachable self-hosted server to issue HTTP or HTTPS requests to internal hosts. A valid image response is returned in Base64 form, while other responses can expose limited timing and error information for internal-network reconnaissance. Exploitation requires a publicly reachable self-hosted instance with public registration enabled, an authenticated AliasVault account, an attacker-controlled authoritative DNS server, and a resolver that permits rebinding; default checks limit requests to ports 80 and 443. The browser extensions, web application, and mobile applications are not affected. This issue is fixed in version 0.30.3.
-
CVE-2026-94620
•
published on October 1, 2026
Classroom 50 is a free and open-source tool for managing and grading programming assignments via GitHub. Prior to version 1.11.0, `gh teacher download` clones each student's assignment repository and then writes autograde artifacts (`result.json` and `results.json`) into the just-cloned working tree. The write followed symlinks, so a student who committed `result.json` or `results.json` as a **symlink** (materialized verbatim by `git clone`) could redirect the teacher's write to an arbitrary path — e.g. `~/.zshrc`, `~/.ssh/authorized_keys`, a cron file, or an in-clone `.git/hooks/*` file that git subsequently executes. The written bytes are attacker-controlled (the student's uploaded release asset for `result.json`; student-chosen submit-tag names for `results.json`). This is an arbitrary file write leading to code execution as the teacher, whose `gh` token carries `admin:org`, `repo`, and `workflow` across the entire classroom organization. Version 1.11.0 contains a patch. Some workarounds are available. Avoid running `gh teacher download` against untrusted student repositories, or run it inside a disposable sandbox / container with no access to sensitive host files or credentials. Inspect cloned trees for symlinked, hardlinked, or special (`result.json`/`results.json`) entries before allowing the artifact-refresh step to run.
-
CVE-2026-94619
•
published on November 6, 2026
FileGator is a multi-user file manager. Prior to 7.15.0, POST /upload in backend/Controllers/UploadController.php uses the request-supplied resumableFilename as an unnamespaced key in shared temporary storage. In a multi-user deployment configured with distinct per-user home directories, a low-privileged user with upload permission can set resumableTotalChunks and resumableTotalSize to zero so the assembly loop writes no data, after which the handler reads a pre-existing temporary file with that name, stores it in the attacker's home directory, and removes it. The same filename path can append data when chunks are supplied, allowing disclosure, modification, or deletion of another user's transient object, including a pending archive or the login rate-limit lockfile, while Tmpfs path sanitization limits the primitive to the temporary directory. This issue is fixed in version 7.15.0.
-
CVE-2026-94618
•
published on November 6, 2026
FileGator is a multi-user file manager. Prior to 7.15.0, GET /batchdownload in backend/Controllers/DownloadController.php uses batchDownloadStart() to stream a shared temporary archive selected by the request-supplied uniqid without checking that the current session created it. In deployments with distinct per-user home directories, a valid user with read, download, and batchdownload permissions who obtains another user's active archive id can cross that boundary and retrieve the victim's batch archive. The archive lifetime and id-acquisition requirement limit exploitation, but a successful replay also deletes the victim's pending archive. This issue is fixed in version 7.15.0.
-
CVE-2026-94613
•
published on September 24, 2026
authentik is an open-source identity provider. Prior to 2026.2.7, 2026.5.7, and 2026.8.2, an unauthenticated attacker can submit a malformed SAML message to an authentik deployment using SAML in either the identity-provider or SAML source role. The message can stop the worker handling /application/saml/* or /source/saml/*, causing the requests assigned to that worker to fail. Worker process termination and automatic restart do not destroy database-backed sessions, but continued malicious messages can cause a sustained share of legitimate traffic to fail. Other protocol implementations are not affected. This issue is fixed in versions 2026.2.7, 2026.5.7, and 2026.8.2.
-
CVE-2026-94612
•
published on September 24, 2026
authentik is an open-source identity provider. Prior to 2026.2.7, 2026.5.7, and 2026.8.2, an authentik SAML Source verifies an assertion's signature and validity period but does not ensure that the identity provider issued the assertion for that Source or in response to a login request from that Source. The SAML Source also does not record already accepted assertions, allowing replay. An unauthenticated actor who possesses such a valid assertion can use an assertion intended for another service provider or reuse an earlier assertion to authenticate as the user named by the assertion. Only SAML Sources are affected; SAML Providers and other Source types are not affected. This issue is fixed in versions 2026.2.7, 2026.5.7, and 2026.8.2.
-
CVE-2026-94611
•
published on September 24, 2026
authentik is an open-source identity provider. Prior to 2026.2.7, 2026.5.7, and 2026.8.2, authentik API serializers return stored credentials when an account has view permission on an affected configuration, even when that account is not authorized to change the configuration or read its secrets. Affected configurations include one-time code delivery by mail or SMS, outbound provisioning targets, device trust integrations, identity sources, the Kubernetes outpost integration, applications using a client or shared secret, and applications using a proxy provider. Deployments are affected when view permission is granted to accounts that are not intended to read these credentials; deployments where every viewer is permitted to read them are not affected. This issue is fixed in versions 2026.2.7, 2026.5.7, and 2026.8.2.
-
CVE-2026-94609
•
published on September 24, 2026
authentik is an open-source identity provider. Prior to 2026.2.7, 2026.5.7, and 2026.8.2, an account with delegated permission to manage a group, group membership, or a user can grant superuser status to an account or assign an existing role to a group without holding the permissions that gate those privileges. Group hierarchy checks do not consistently account for superuser status inherited from ancestor groups, and role assignment to a group lacks the required authorization check. Only deployments that delegate these management capabilities to accounts that are not full administrators are affected. This issue is fixed in versions 2026.2.7, 2026.5.7, and 2026.8.2.
-
CVE-2026-94606
•
published on September 24, 2026
authentik is an open-source identity provider. Prior to 2026.2.7, 2026.5.7, and 2026.8.2, authentik email authenticator enrollment during an authentication or enrollment flow accepts a recipient address supplied in the setup request instead of using the address already established by the flow. An actor who knows a target user's password can substitute an attacker-controlled address, receive the one-time code, and finish enrolling the factor as the target. The target must not have enrolled the email factor already. Successful enrollment gives the actor a session as the target and access to single sign-on applications behind the account. Other authenticator types are not affected. This issue is fixed in versions 2026.2.7, 2026.5.7, and 2026.8.2.
-
CVE-2026-94603
•
published on October 13, 2026
Podman is a tool for managing OCI containers and pods. From 4.4.0 until 5.8.8 and 6.1.3, podman run treats an image carrying the io.podman.annotations.checkpoint.runtime.name annotation as a checkpoint and silently ignores user-supplied container creation restrictions such as --cap-drop=ALL, allowing image-controlled checkpoint state to select weaker sandbox settings and capabilities. Rootless containers are not affected because checkpoint and restore are unavailable in that mode. This issue is fixed in versions 5.8.8 and 6.1.3.