| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| vLLM is an inference and serving engine for large language models. Prior to 0.24.0, the input_audio handling path for /v1/chat/completions calls AudioMediaIO.load_bytes or AudioMediaIO.load_file without passing VLLM_MAX_AUDIO_DECODE_DURATION_S to the shared audio decoder. An unauthenticated client can therefore submit a small compressed audio input that expands into a very large float32 PCM allocation, bypassing the duration guard already used by /v1/audio/transcriptions and causing an out-of-memory worker crash. Inline data URLs reach this path without being bounded by VLLM_AUDIO_FETCH_TIMEOUT. The issue affects deployments serving an audio-capable model, and authentication changes only the deployment-specific reachability. This issue is fixed in version 0.24.0. |
| In stpropnci_process of stpropnci.cc, there is a possible out of bounds write due to an incorrect bounds check. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation. |
| vLLM is an inference and serving engine for large language models. Prior to 0.28.0, request bodies for Chat Completions and Responses can set media_io_kwargs.video.video_backend to pynvvideocodec, and MediaConnector.fetch_video forwards that choice to VideoMediaIO even when startup configuration selected a software decoder. The engine's _reserve_mm_ipc_gpu_memory logic budgets decoder memory only from static configuration, so the request-selected VIDEO_LOADER_REGISTRY backend can create a CUDA context, decoder surfaces, and decoded-frame allocations that were not removed from the engine's KV-cache budget. An attacker able to submit video requests to a video-capable GPU deployment with PyNvVideoCodec installed can exhaust shared GPU memory, causing request failures, worker crashes, or denial of service. The first release containing the fix is version 0.28.0. |
| In the Linux kernel, the following vulnerability has been resolved:
RDMA/erdma: Use IRQ-safe XArray helpers for QP and CQ tables
Locked QP and CQ lookups from EQ interrupts can deadlock with
create-path XArray updates. If an interrupt arrives while the create
path holds the plain xa_lock, the lookup spins forever trying to
acquire the same lock.
Use IRQ-safe XArray helpers for all QP and CQ create-path updates,
including the GSI QP store and error paths. Initialize both arrays with
XA_FLAGS_LOCK_IRQ so sleeping allocations preserve interrupt state. |
| A flaw has been found in XnView Classic 2.52.5. Impacted is an unknown function of the component FLI File Parser. This manipulation of the argument starting_line causes heap-based buffer overflow. Remote exploitation of the attack is possible. Upgrading to version 2.52.6 is recommended to address this issue. Upgrading the affected component is advised. |
| Improper neutralization of special elements used in a command ('command injection') in Microsoft 365 Copilot's Business Chat allows an unauthorized attacker to disclose information over a network. |
| A stored cross-site scripting (XSS) weakness in Kiteworks Core could allow an administrator holding only a single, narrowly scoped delegated permission to store crafted content that later executes arbitrary JavaScript in the authenticated session of a System Administrator who views the affected page. This could have permitted the lower-privileged administrator to escalate to full administrative control of the tenant, including the creation of a new administrative account. |
| MultiversX's multisig-improved (repository: mx-multisig-and-modules) reference implementation of their on-chain multisig smart contract system contains a vulnerability where a missing independent authorization check allows any account with the Proposer role to perform explicitly barred actions. This vulnerability allows the Proposer role to move funds alone, draining 100% of a contract's EGLD/ESDT balance in two transactions with zero signatures. |
| An administrative import function in Kiteworks Core did not verify that the requesting administrator was entitled to create the privileged integration credential being imported. A delegated administrator holding a single narrowly scoped administrative permission could therefore obtain full system administrator privileges, without any action by an existing system administrator. |
| Kiteworks Core did not apply its gateway-level API security controls to every request authenticated through the platform's central authentication service. An authenticated user could reach REST API functionality over a request path on which those controls, including enforcement of signed-out and revoked sessions, were not applied. |
| A Kiteworks appliance setup interface did not confine a user-supplied file path to its intended directory, which could allow an unauthenticated attacker to write a file to any location writable by the affected service account, potentially compromising the integrity of the appliance or rendering it unavailable until an operator intervenes. Exploitation requires network access to the affected interface, which is not reachable on a fully configured appliance in its default configuration; reaching it depends on either the transient window while an appliance is first being provisioned or a non-default appliance configuration. |
| In NTFS-3G before 2026.7.7, a heap buffer overflow exists in ntfs_index_walk_down() in libntfs-3g/index.c that allows an attacker to corrupt heap memory in the SUID-root ntfs-3g binary by crafting a malicious NTFS image. The overflow is triggered by reading the special crafted file metadata. |
| Kiteworks did not correctly enforce which roles a shared folder's manager was permitted to assign. In a default configuration, an authenticated user holding the Manager role on a folder could grant the Owner role to themselves or to other members of that folder. |
| In NTFS-3G before 2026.7.7, a out-of-bounds read exists in ntfs_fix_file_name() in libntfs-3g/reparse.c that allows an attacker to read possibly confidential information in ntfs-3g process memory by crafting a malicious NTFS image. The out-of-bounds read is triggered by a readlink on a corrupted file. |
| All versions of Zammad including the latest alpha enable the local zammad user to escalate privileges to root. |
| Zammad versions 6.3.0 to 6.5.4 are vulnerable a session hijack vulnerability that leads to remote code execution as the zammad user. The bug is also present in version 7.0.0 to version 7.1.2, but not exploitable due to changes in the underlying framework. |
| In NTFS-3G before 2026.7.7, a heap buffer overflow exists in ntfs_ib_copy_tail(), in libntfs-3g/index.c, that allows an attacker to corrupt heap memory in the SUID-root ntfs-3g binary by crafting a malicious NTFS image. The overflow is triggered by extending a directory, e.g., by creating a file. |
| Rundeck before 6.2.0 contains an OS command injection vulnerability that allows authenticated users with job run permission to execute commands on Windows nodes by supplying crafted option values. Attackers can inject cmd.exe metacharacters such as && or | into free-text options, which CLIUtils.quoteWindowsCMDArg wraps in ineffective single quotes, running commands with node executor credential privileges. |
| wger is a free, open-source workout and fitness manager. Prior to version 2.6, an authenticated attacker can inject arbitrary workout log entries into any other user's `SlotEntry` by supplying the victim's `slot_entry` ID in a `POST /api/v2/workoutlog/` request. The `slot_entry` foreign key is not included in the ownership verification performed by `WorkoutLogViewSet.get_owner_objects()`, so the server accepts and persists the cross-user reference without error. Because `SlotEntry.get_config_data()` retrieves associated logs via `self.workoutlog_set.all()` with no user filter, the attacker's injected data is silently folded into the victim's progressive-overload calculations, corrupting their auto-generated weight and repetition targets. Version 2.6 contains a patch. |
| In NTFS-3G before 2026.7.7, a heap buffer overflow exists in ntfs_ir_to_ib() in index.c that allows an attacker to corrupt heap memory in the SUID-root ntfs-3g binary by crafting a malicious NTFS image. The overflow is triggered by extending a directory, e.g., by creating a file. |
