LibreOffice Writer can save documents in DOCX format. A document can contain content controls whose values are bound to custom XML data, and that data is updated with the current values when the document is exported, including in a conversion run from the command line. Values taken from the document were not escaped where they were used, so a crafted document could write a file of its choosing anywhere the user can write when it was exported to DOCX. In fixed versions these values are written into the document directly, and the XSLT transformations LibreOffice runs can no longer write files or use the network.
LibreOffice documents in ODF format can contain form controls, such as buttons and text fields, and each control names the component that implements it. The component a control named was created even when it was not a form control, and the settings stored with the control were applied to it, so a crafted document could change the configuration of LibreOffice in a way that led to code being run. In fixed versions a control whose component is not a form control is ignored.
CVE-2025-43874 is an OS Command Injection vulnerability affecting multiple Johnson Controls iSTAR Ultra product lines prior to version 6.9.2.CU02. The flaw involves improper neutralization of special elements in OS commands, which could allow an attacker with limited privileges to execute arbitrary OS commands. The vulnerability has a high severity score of 8.7 but no known exploits in the wild have been reported. A fixed version 6.9.2.CU02 or later is indicated, though no explicit patch link is provided.
CVE-2026-116311 is a vulnerability in PostgreSQL where the DROP OWNED command can remove restrictive row security policies from tables that the user does not own. This behavior could lead to unintended removal of security policies, potentially weakening data access controls.
A path parsing flaw in Nuxt versions 4.6.0 through 4.6.3 allows an unauthenticated attacker to bypass server-side route controls by using request paths starting with double slashes (//). This causes server middleware and route rules keyed on normal paths (e.g., /admin) to be skipped, enabling unauthorized access to protected pages and skipping security headers or redirects. The issue is fixed in version 4.6.4.
OpenBao versions before 2.6.5 and 2.7.x before 2.7.2 have a vulnerability where wildcard characters in SSH certificate principals are not neutralized. This occurs when principals are rendered from unvalidated external token metadata via identity templates in the SSH secrets engine. An attacker who controls this metadata, often through a compromised identity provider, can inject '*' characters to obtain SSH certificates valid for multiple hosts or users.
CVE-2026-126192 is a performance-related vulnerability in Soup Sieve where using a general sibling combinator (~) inside the :has() CSS selector causes quadratic time complexity when matching against untrusted HTML. This can lead to significant CPU consumption and potential denial of service if an attacker controls the HTML input and the selector is applied. The issue arises because the selector implementation performs nested scans over sibling elements, causing processing time to grow quadratically with the number of siblings. The vulnerability affects Soup Sieve versions starting from 1.3.0. Mitigations include careful use of :has() selectors, avoiding large numbers of children on the anchor element, and preferring more restrictive combinators like > or + inside :has().
The wasmtime-wasi-http p2 implementation in Wasmtime before 36.0.20, 37.0.0 before 48.0.8, 49.0.0 before 49.0.5, and 50.0.0 before 50.0.1 can send conflicting content-length headers, or ones not matching the body, in outgoing requests. A guest that reuses a received response's headers resource to send a bodyless request can enable request smuggling against non-compliant servers, potentially bypassing embedder destination access controls.
MCUboot's Zephyr serial-recovery upload handler bs_upload() in boot/boot_serial/src/boot_serial.c decodes the host-supplied image field of an MCUmgr image-upload request into img_num (a full 32-bit value) and latches it on the off == 0 packet without any range check. It then passes img_num straight to flash_area_id_from_multi_image_slot(img_num, 0), which on the Zephyr port resolves to __flash_area_ids_for_slot() in boot/zephyr/include/sysflash/sysflash.h and indexes the static const int all_slots[] table as all_slots[img_num 2]. That table holds only MCUBOOT_IMAGE_NUMBER 2 entries, so any img_num >= MCUBOOT_IMAGE_NUMBER is an out-of-bounds read.
The path is reachable by anyone with access to the serial / SMP recovery console, which is unauthenticated by design, and the bad value is supplied before any image signature is verified. The defect is specific to the default non-direct upload path (MCUBOOT_SERIAL_DIRECT_IMAGE_UPLOAD disabled) on the Zephyr port with the array-backed slot layout (not CONFIG_SINGLE_APPLICATION_SLOT). It is not gated on a multi-image configuration: MCUBOOT_IMAGE_NUMBER defaults to 1, so even a default single-image build is affected for any image >= 1. The MCUBOOT_SERIAL_DIRECT_IMAGE_UPLOAD path is safe (its flash_area_id_from_direct_image() switch returns -EINVAL for unknown indices).
flash_area_id_from_multi_image_slot() takes a signed int image_index while img_num is uint32_t, so the attacker controls the sign of the index as well as its magnitude: an image of 0xFFFFFFFE converts to -2 and reads all_slots[-4], off the front of the table, while large positive values read off the end (CWE-192). (0xFFFFFFFF itself is the "no image given" sentinel and selects image 0.) Either way the index multiply img * 2 is performed in int, so a wild offset yields a bus/hard fault, giving a reliable denial of service of the bootloader during recovery. For moderately out-of-range indices the read returns adjacent .rodata as a garbage flash-area id that is handed to flash_area_open(); this normally fails with MGMT_ERR_EINVAL, but if the garbage id happens to match a valid non-image partition, the subsequent erase/write can target a partition outside the image set (CWE-787). The attacker-supplied value is never echoed back, so there is no information disclosure. The fix is to reject img_num >= BOOT_IMAGE_NUMBER on the off == 0 packet, before the slot lookup. The comparison must be made on the unsigned img_num: an equivalent guard applied after the conversion to int would not reject the negative case above. __flash_area_ids_for_slot() should bounds-check its own arguments as defence in depth.
This vulnerability was reported independently twice: by NVIDIA Project Vanessa on 2026-06-22, and by Kai Phan (@kaimandalic) on 2026-07-19 as GHSA-8v33-88p6-jffj, which is closed as a duplicate of this advisory. Both original reports are reproduced in full below.
A critical security vulnerability (CVE-2026-71195) affects the Red Hat OpenStack Platform 17.1, specifically the openstack-manila component. This flaw allows unauthorized cross-project access-rule and export-metadata writes on public Manila shares, potentially bypassing security controls in a multi-tenant cloud environment. Red Hat has released a security update to address this issue.
Addressable versions from 2.0.0 up to 2.9.1 contain a vulnerability in the IDNA.to_unicode method that allows decoding of Punycode labels with bidirectional controls, zero-width characters, control characters, or other default-ignorable format characters. This can cause the display_uri method to show reordered or visually indistinguishable hostnames compared to the actual encoded hostname used for network connections. The issue affects both the pure-Ruby backend and the native libidn backend (from 2.5.2 for certain Unicode characters). The vulnerability is fixed in version 2.9.1.
An Insecure Direct Object Reference (IDOR) vulnerability exists in the Joomla extension jDownloads versions prior to 4.1.8. The vulnerability allows unauthorized users to download restricted files by manipulating the 'other_file_id' parameter, bypassing access controls intended to protect those files.
@hono/node-server allows running the Hono application on Node.js. Prior to 1.19.18 and from 2.0.0 until 2.1.6, the server passes a raw request-target containing a backslash in its path to the WHATWG URL parser, which treats the backslash as a path separator and resolves dot segments. A reverse proxy, load balancer, or web application firewall that preserves the backslash and applies path-based rules to the original request can therefore permit a request that the application normalizes to a blocked route. This lets an unauthenticated client reach route content that the upstream component was intended to deny, while Hono application access controls are unaffected because they operate on the normalized path. This issue is fixed in versions 1.19.18 and 2.1.6.
A vulnerability in AWX allows unauthenticated remote attackers to access debug API endpoints that trigger scheduler operations acquiring database advisory locks and opening transactions. This can lead to resource exhaustion and degraded service availability due to lock contention. The flaw arises because these debug endpoints are registered unconditionally and use permissive access controls. While the vulnerability does not allow data compromise, it can cause service degradation under sustained attack. No complete mitigation exists except applying a future update. Interim mitigations include restricting network access and blocking debug API paths.
Two studies by Microsoft reveal post-training supply chain attacks on AI coding agents through dataset poisoning and model poisoning. Dataset poisoning involves inserting crafted examples into fine-tuning data that teach the model to misuse available command-execution tools when triggered by benign-looking user requests. Model poisoning assumes attacker control over training and distribution of model checkpoints, embedding reliable backdoors that activate unsafe behaviors while maintaining general performance. These backdoors can cause agents to execute commands, modify files, or access data without obvious malicious triggers, posing risks to sensitive development environments. The attacks exploit trust in training data and model artifacts, emphasizing the need for controls over training pipelines, model provenance, and tool permissions.
The SANSAR Linux cryptomining intrusion is a sophisticated malware operation that uses a Rust-based orchestrator to manage CPU and GPU miners via Tor. Attackers exploited an exposed Ray job agent to gain root access on Linux GPU hosts, deploying a kernel rootkit to hide processes and files. It includes an NVIDIA Management Library wrapper that falsifies GPU telemetry to show 0% utilization, concealing cryptomining activity. The malware also uses a backdoor named netns to retrieve command and control addresses from Ethereum smart contracts. SANSAR controls XMRig and SRBMiner-MULTI miners and maintains command handlers for configuration and interactive shells. Multiple stealth techniques such as ftrace-based rootkit hooks, modified GPU monitoring libraries, and Tor-based communications with bridge support enhance anonymity and evasion.
An incorrect authorization vulnerability was found in AWX's WebSocket consumer. The consumer_access() function only enforces access controls for object-level event groups (job_events, workflow_events, ad_hoc_command_events) but does not restrict subscriptions to global broadcast groups such as 'jobs-status_changed'. When an authenticated user subscribes to a global broadcast group, the server accepts the subscription without verifying that the user has permission to view the associated resources. In multi-organization deployments, this allows a low-privilege user in one organization to receive real-time job status metadata from all organizations, including job IDs, status transitions, timestamps, and infrastructure topology details such as instance group names, bypassing the organization isolation enforced by the REST API.
The MISP Modules website component has a vulnerability in its external tools management endpoints that allows unauthenticated attackers to add, delete, modify, and toggle external tools without authorization. This occurs because the external tools handlers do not enforce the administrator authentication policy even when ADMIN_PASSWORD is set. The flaw affects the external_tools Flask blueprint and enables bypass of intended access controls. Installations without an admin password are designed to allow open management access, so the vulnerability is conditional on ADMIN_PASSWORD being configured.