CVE-2026-105140 is a race condition vulnerability in obot-platform's obot product affecting versions 0.25.0 before 0.25.6 and 0.26.0 before 0.26.1. The flaw occurs during concurrent auth provider group refreshes, allowing revoked group memberships to be temporarily restored due to out-of-order commits. This results in users retaining revoked group-based access for approximately ten minutes.
British online retailer ASOS confirmed a cyberattack involving the compromise of a third-party communication platform used to send notifications to users. Hackers sent rogue notifications to ASOS users' mobile apps, claiming full compromise of ASOS's Snowflake data platform instance and threatening data leaks. ASOS stated that basic user information such as names and contact details may have been accessed, but payment card information and passwords were not impacted. The company’s website and app operations remain unaffected. The incident was claimed by a hacking group named Xuanye Group, possibly a Chinese-speaking threat actor or a false flag. ASOS has restricted access to the compromised notification platform and is working with specialists and authorities. The exact initial access vector and details about the third-party platform remain undisclosed.
Apache YuniKorn 1.8.0 and later, if configured with the LDAP group resolver, crashes due to an out of bounds read processing group membership entries.If the LDAP server returns a group membership entry, memberOf attribute, for a user specified in the pod the server crashes if a membership record does not start with "CN=".
This only affects install that have the non default LDAP group provider configured.
Users are recommended to upgrade to version 1.10.0, which fixes this issue.
src/internal.c in wolfSSL wolfSSH through 1.5.0 admits the server-to-client Diffie-Hellman group exchange messages SSH_MSG_KEX_DH_GEX_GROUP (31) and SSH_MSG_KEX_DH_GEX_REPLY (33) when a server receives them from an unauthenticated client. IsMessageAllowedServer() applies no direction check to the key exchange message range: when the peer is keying and no particular message is expected, which is the state a server is in for the whole window after it processes the client's KEXINIT because nothing sets handshake->expectMsgId there, the function falls out of its expectation branch without a verdict and reaches a numeric bound that admits every message id from 30 through 34. A client that negotiates diffie-hellman-group-exchange-sha256 and then sends message 31 makes the server run the client-side handler DoKexDhGexGroup(), which validates the attacker-supplied group with two 8-round Miller-Rabin primality tests, one on p and one on (p-1)/2, on a value of up to 8192 bits. The handler then returns success: the server stores the attacker's prime and generator, generates a Diffie-Hellman key pair in the attacker's group, and sends the client-role message SSH_MSG_KEX_DH_GEX_INIT (32) back to the attacker. Published RFC 3526 safe primes are the worst-case input and cost the attacker nothing to obtain. The primality validation was added in 1.5.0; versions from 1.2.0 through 1.4.22 admit the same message and enter the same client-role path without the primality cost. Message 33 is admitted as well, but on a server it is rejected before any cryptography because no public key check callback is registered, so it carries no comparable cost. Builds that define WOLFSSH_NO_DH_GEX_SHA256, which is implied by WOLFSSH_NO_DH or NO_SHA256, are unaffected.
CVE-2026-106547 is a heap-based buffer overflow vulnerability in the HDF5 library before version 2.2.0. It occurs in the H5VM_array_fill() function when reading unallocated chunks of a dataset with inconsistent metadata, allowing writes beyond the allocated buffer size. This flaw can cause application crashes and potentially arbitrary code execution when processing crafted HDF5 files.
CVE-2026-19029 is a heap-based buffer over-read vulnerability in the HDF5 library (The HDF Group) affecting versions prior to 2.2.0. The flaw exists in the H5Z__filter_scaleoffset() function, where an attacker can craft a malicious HDF5 file that triggers a denial of service via application crash. This occurs because the decoder copies data from a compressed chunk without verifying the chunk size when certain filter parameters are attacker-controlled.
CVE-2026-104048 is a vulnerability in SSSD on Red Hat Enterprise Linux 10 where Host-Based Access Control (HBAC) rules are improperly evaluated in trust-enabled identity management environments. The flaw occurs because SSSD strips domain qualifiers and compares only short usernames, allowing an authenticated user in a trusted domain with the same username as a local authorized account to bypass access policies. Exploitation requires a non-default trust deployment and username collision. This vulnerability cannot be exploited remotely by unauthenticated users and does not affect environments without cross-realm trusts or those using group-based HBAC policies.
ImageSharp is a 2D graphics library. From 3.0.0 until 4.1.1, decoding a strip TIFF using CCITT Group 3 or Modified Huffman compression can pass attacker-expanded runs to BitWriterUtils.WriteBits without first checking the current row width. T4TiffCompression.WritePixelRun can accumulate oversized makeup-code runs, and ModifiedHuffmanTiffCompression.Decompress validates the width only after writing. The unchecked writes can overflow the strip buffer, corrupt heap memory, and terminate the process. This strip-path vulnerability is distinct from the tiled decompressor-width mismatch. This issue is fixed in version 4.1.1.
SimpleChat is a secure AI conversation application with personal and group workspaces for document-grounded interactions. Prior to 0.261.029, POST /api/group_documents/upload stores an attacker-controlled group document filename that group_workspaces.html later interpolates into inline Share event handlers. The escapeGroupHtml function leaves apostrophes unchanged, while escapeHtml produces an HTML entity that the browser decodes before JavaScript evaluation, so either path permits the filename to terminate the handler string. An authenticated group Owner, Admin, or DocumentManager can persist script that executes in the SimpleChat origin when another group member clicks Share, allowing access to victim-visible data and actions with the victim session. This issue is fixed in version 0.261.029.
SimpleChat is a secure AI conversation application with personal and group workspaces for document-grounded interactions. In versions 0.261.003 and 0.261.027, an authorization ordering flaw in POST /api/user/plugins allows an authenticated low-privileged user to omit the top-level MCP type so that _reject_non_admin_mcp_stdio skips inspection before the type is restored from metadata. The stored personal action can then reach McpPluginFactory.create_connector, and MCPStdioPlugin.connect starts the attacker-selected operating-system process under the application service identity when the action tool is invoked. Exploitation requires personal plugins to be enabled and governance to permit MCP actions, and it can expose or modify secrets and data available to the service or disrupt the service. This issue is fixed in version 0.261.031.
The FBI experienced a data breach exposing personal information of thousands of its employees due to a missed security patch on an Oracle PeopleSoft platform managed by an Accenture contractor. The breach was attributed to the ShinyHunters cybercrime group, which exploited the unpatched system to access sensitive data. The FBI removed the contractor responsible and took measures to mitigate further risk. The incident followed ShinyHunters' claim of hacking the FBI's job site to pressure the agency over a report about the group. Law enforcement has arrested alleged leaders of ShinyHunters in connection with the breach.
CVE-2026-98360 is a high-severity vulnerability in the Linux kernel's RDMA/rxe subsystem. It involves improper handling of multicast group insertion into the mcg_tree before the backing Ethernet multicast address is programmed. This can lead to use-after-free errors when a multicast group is freed without being removed from the tree after a failure in rxe_mcast_add(). The issue has been fixed by ensuring the multicast group is only published after successful programming of the multicast address.
A vulnerability in the Linux kernel's wifi cfg80211 subsystem involved improper grouping of hidden BSS entries without beacon elements. This could cause internal warnings due to unexpected grouping behavior when probe responses and beacons are processed. The issue has been resolved by ensuring only entries with beacon information are grouped, preventing the problematic state.
Insufficient verification of data authenticity (CWE-345) in the external OIDC login callback in Cloud Foundry UAA v4.5.0 to v79.6.0 (inclusive) allows an authenticated UAA user to bypass the OAuth authorization-code exchange and establish an authenticated external-OIDC browser session, via submitting a UAA access token or a cross-client ID token as the callback’s id_token parameter.
The issue only manifests when a UAA zone is configured with an OIDC identity provider whose issuer exactly matches that zone’s own /oauth/token endpoint (a “self-UAA” OIDC configuration). In this configuration, the callback takes a supplied id_token directly instead of requiring the authorization code exchange, and does not verify that the token was actually issued as an ID token for the specific self-OIDC relying-party client. An attacker holding any valid UAA JWT for themselves — including a plain access token with only uaa.user scope, or a valid ID token issued to an unrelated client such as cf — can present it as the callback’s id_token and be authenticated into a mapped local (“shadow”) account. Because the resulting session is not verified against the originating token’s true audience or user_id, its effective privilege depends entirely on the shadow account’s group memberships, which can include administrative scopes such as clients.write.
Exploitation requires a valid UAA user JWT, a valid browser login state for the target zone, and the presence of a self-referential OIDC provider configuration — this is not a pre-authentication vulnerability, and does not by itself grant privileges beyond those already held by the mapped shadow account.
A Russian-speaking affiliate of the Gentlemen ransomware group, known as Azazel, operated a sophisticated ransomware and data exfiltration campaign targeting over two dozen organizations across six countries. Azazel exploited stolen CI/CD secrets, GitLab variable stores, and git history to gain initial access, and used novel techniques such as leveraging an AI coding assistant with a reverse shell over MCP as a command and control channel. The operation involved more than 50TB of dedicated bare-metal infrastructure and approximately 6TB of stolen data. Azazel ran an independent leak site to publish victim data and collect ransom payments, bypassing the main ransomware-as-a-service operator. The investigation uncovered active exfiltration during the analysis and detailed victim notification was conducted prior to public disclosure.
CVE-2026-104043 is a vulnerability in SSSD on Red Hat Enterprise Linux 10 where a local attacker with access to the NSS responder UNIX socket can cause an integer underflow by sending a malformed request. This triggers an out-of-bounds memory read during packet parsing, crashing the responder process and causing a denial of service (DoS). The flaw does not allow privilege escalation or remote code execution but disrupts identity and group resolution on the host.
CVE-2026-104036 is a vulnerability in the SSSD NFS idmap plugin of Red Hat Enterprise Linux 10. It involves an out-of-bounds write when retrieving cached user or group names due to improper handling of oversized entries. This flaw can be triggered by a local attacker with low privileges, causing a denial of service by crashing the identity mapping service and potentially corrupting adjacent process memory. Exploitation requires specific conditions, including the use of rpc.idmapd with the SSSD NFS plugin and memcached lookups enabled. No official patch is currently available, but a configuration workaround is recommended.
### Summary
`allow_external_plugins=False` (the default, and the CLI default) is meant to restrict docling to its own model plugins. However, docling's plugin factories call pluggy's `load_setuptools_entrypoints()`, which imports every module registered under docling's plugin entry-point group. Only afterwards does docling filter out modules outside the `docling.` namespace. Import-time code in any installed third-party plugin therefore runs even though external plugins are disabled.
### Details
In `docling/models/factories/base_factory.py`, `load_from_plugins()` loads all entry points first and applies the `allow_external_plugins` check only to the already-imported modules. The CLI creates these factories when it starts, so running `docling` imports every registered plugin module. A log message says the plugin "will not be loaded", although its module has already been imported.
### Affected configurations
Environments in which a package registering a docling plugin entry point is installed, for example an unvetted or compromised dependency, and which rely on `allow_external_plugins=False` to keep that code from running.
### Impact
Execution of a third-party plugin module's import-time code in the docling process, contrary to the documented behaviour of `allow_external_plugins=False`.
### Patches
Fixed in docling 2.131.0 by [#4413](https://github.com/docling-project/docling/pull/4413). Plugin entry points are now filtered by module name before they are loaded, so with `allow_external_plugins=False` third-party plugin modules are no longer imported.
### Workarounds
Upgrade to 2.131.0. For older versions:
Only install trusted packages in environments that run docling. Check which packages register docling plugin entry points with `importlib.metadata.entry_points()`.