CVE Database

FlashAttention through 2.8.3.post1, fixed in commit 0816ef1, contains a symlink attack vulnerability in the download_and_copy() function within hopper/setup.py that extracts NVIDIA toolchain archives without validating symlinks or filtering tar members. A local attacker can pre-plant a symlink in the predictable cache directory to redirect extracted binaries to an attacker-chosen location, enabling arbitrary file write with victim privileges during build time.

OpenClaw versions 2026.5.28 before 2026.6.6 contain an authorization bypass vulnerability in native web search that allows lower-trust callers to perform actions requiring stronger policy checks. Attackers can exploit misconfigured input paths to bypass intended authorization controls and execute restricted operations.

OpenClaw versions 2026.6.5 before 2026.6.9 contain a vulnerability in the plugin install wrappers that could skip the install policy (authorization) check. When the affected feature is enabled and reachable, a lower-trust caller or a configured input path could execute or persist actions beyond the caller's intended authorization. Impact depends on the operator's configuration and whether lower-trust input can reach the affected path. The issue is fixed in 2026.6.9.

HedgeDoc is an open source, real-time, collaborative, markdown notes application. Versions prior to 1.11.0 allowed attackers to circumvent the rate-limiting of the /login and /register routes by spoofing IP addresses. HedgeDoc instances checked for CloudFlare's cf-connecting-ip header and used that instead of the users real IP address, if the header was present even when the request did not originate from Cloudflare. This made it possible for an attacker to spam login requests or create multiple arbitrary accounts by sending another cf-connecting-ip header every few requests. The issue has been fixed in version 1.11.0.

HedgeDoc is an open source, real-time, collaborative, markdown notes application. Prior to version 1.11.0, due to unsafe handling of the local-part of registered email addresses, HedgeDoc was vulnerable to stored HTML Injection through its publish and slide views. An attacker could register a specially crafted email address and inject arbitrary HTML into pages viewed by other users. HedgeDoc accepted RFC 5321 quoted-string local-parts in email addresses during registration. The local-part was then reused as the user's display name without escaping and rendered into HTML in multiple places, including publish and slide views as well as the collaborative editor. An attacker could break out of an HTML attribute and inject arbitrary markup into the page. While the deployed Content-Security-Policy prevented straightforward inline JavaScript execution, the injected HTML was still sufficient to alter page content and embed attacker-controlled resources such as cross-origin iframes. This issue was fixed in version 1.11.0.

The SCORM lab launch endpoint in Skillable (scorm.skillable.com) through 2026-07-13 does not validate the client-supplied userId parameter against the authenticated SCORM session token. An authenticated user can substitute arbitrary userId values to bypass per-user lab launch rate limits and consume other users' lab allocations, resulting in denial of service against targeted users' lab and exam access. Skillable was formerly named Learn on Demand Systems.

AnyDesk Support Information Link Following Denial-of-Service Vulnerability. This vulnerability allows local attackers to create a denial-of-service condition on affected installations of AnyDesk. An attacker must first obtain the ability to execute low-privileged code on the target system in order to exploit this vulnerability. The specific flaw exists within the Send Support Information feature. By creating a junction, an attacker can abuse the service to create arbitrary files. An attacker can leverage this vulnerability to create a denial-of-service condition on the system. Was ZDI-CAN-26645.

AnyDesk Screen Recording Link Following Denial-of-Service Vulnerability. This vulnerability allows local attackers to create a denial-of-service condition on affected installations of AnyDesk. An attacker must first obtain the ability to execute low-privileged code on the target system in order to exploit this vulnerability. The specific flaw exists within the handling of screen recording files. By creating a junction, an attacker can abuse the service to create arbitrary files. An attacker can leverage this vulnerability to create a denial-of-service condition on the system. Was ZDI-CAN-26591.

A weakness has been identified in antv layout 2.0.0. This impacts the function setNestedValue in the library lib/util/object.js. Executing a manipulation of the argument path can lead to improperly controlled modification of object prototype attributes. The attack can be launched remotely. The project was informed of the problem early through an issue report but has not responded yet.

A vulnerability was identified in SourceCodester Class and Exam Timetabling System 1.0. The impacted element is an unknown function of the file /subject.php. Such manipulation of the argument subject leads to cross site scripting. It is possible to launch the attack remotely. The exploit is publicly available and might be used.

A vulnerability was determined in SourceCodester Class and Exam Timetabling System 1.0. The affected element is an unknown function of the file /forsubject.php. This manipulation of the argument subject causes cross site scripting. It is possible to initiate the attack remotely. The exploit has been publicly disclosed and may be utilized.

ChurchCRM is an open-source church management system. Prior to version 7.4.0, a low-privileged user can bypass the /admin/export UI and exfiltrate the entire member directory. The POST /CSVCreateFile.php endpoint generates and streams a CSV containing the full Personally Identifiable Information (PII) of every Person/Family record in the database, without performing any feature-level or object-level authorization check beyond the coarse "has any admin permission" gate inherited from the legacy page bootstrap. In other words, any single non-admin permission flag is enough to reach the CSV bulk-export endpoint, even though such users should not have data export rights. The export script is missing a dedicated isAdmin() (or a new bExportData) authorization check of its own. This issue has been fixed in version 7.4.0.

The Avada (Fusion) Builder plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the ‘Module Title’ parameter in all versions up to, and including, 3.15.5 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with Contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page.

The Smart Slider 3 plugin for WordPress is vulnerable to Sensitive Information Exposure in all versions up to, and including, 3.5.1.37 via the 'keyword' parameter. This makes it possible for authenticated attackers, with contributor-level access and above, to extract titles and full content excerpts of private, draft, pending, trashed, and auto-draft posts authored by any user, including Administrators and Editors. The required nonce is emitted on /wp-admin/post-new.php, which is accessible to Contributor-level users via the edit_posts capability, meaning any Contributor can obtain the nonce needed to trigger the injection.

Cross-site scripting vulnerability in phoenixframework phoenix_live_view allows an attacker to bypass URL scheme validation and execute JavaScript in a victim's browser session. The Phoenix.LiveView.Utils.valid_destination!/2 and Phoenix.LiveView.Utils.valid_live_navigation_destination!/2 functions in lib/phoenix_live_view/utils.ex rely on an internal uri_scheme/1 helper that only detects a scheme when the input's first byte is an ASCII letter. Inputs beginning with an ASCII control character or space fall through to a nil-returning clause, causing the URL to be treated as a safe relative path. Standard browsers implement the WHATWG URL parser, which strips leading C0 control and space characters before parsing. As a result, an input such as " javascript:alert(1)" is passed unchanged into <.link href={...}> and, when clicked, is parsed by the browser as a javascript: URL that executes attacker-controlled script in the victim's session. Applications that render user-supplied URLs (for example profile links, redirect targets, or external references) via <.link href={...}> are affected. This issue affects phoenix_live_view: from 1.2.2 before 1.2.7.

Laravel-Mediable before 7.0.0 contains a stored cross-site scripting vulnerability that allows authenticated or anonymous users to execute arbitrary JavaScript by uploading unsanitized SVG files containing embedded scripts in onload event handlers, script tags, or foreignObject elements. Attackers can store persistent XSS payloads in uploaded SVG files that execute with full DOM access when victims open or preview the file, enabling session cookie theft, CSRF token capture, and account takeover.

Pages with malicious titles could potentially allow saved PDF content to overwrite PDF files or bundled content within the Firefox for iOS application sandbox. This vulnerability was fixed in Firefox for iOS 152.4.

Rejetto HFS 3.0.0 through 3.2.0 allows path traversal through the lang query parameter, permitting a remote unauthenticated attacker to read certain JSON files outside the shared folders. Exploitation is constrained to files matching a narrow naming and format pattern, limiting practical impact.

Rejetto HFS 3.0.0 through 3.2.0 does not escape file names in its fallback "basic" web listing, and this listing can be forced by any browser via the ?get=basic parameter. A user with upload permission - or an anonymous user on servers with an open upload folder - can store a file whose name contains script that executes in the browser of anyone viewing the listing.

Rejetto HFS 3.0.0 through 3.2.0 returns observably different responses from its login endpoint depending on whether the submitted username exists. A remote unauthenticated attacker can use this to confirm valid account names, including the default admin account, facilitating password-guessing and session-forgery attacks.

Rejetto HFS 3.0.0 through 3.2.0 accepts state-changing API requests via the GET method and exempts GET requests from its anti-CSRF header check. A remote attacker can perform administrative actions including account creation and configuration changes leading to code execution - by causing a logged-in administrator's browser to navigate to a crafted URL, or without any credentials against default installations when the attack originates from the server's own machine.

Rejetto HFS 3.0.0 through 3.2.0 renders log entries in the administration panel as HTML without sanitization. A remote unauthenticated attacker can submit a failed login with a crafted username that is written to the error log and executes JavaScript in an administrator's browser when the logs are viewed, allowing the attacker to create accounts or execute code on the server with the administrator's privileges.

Blender 3.0.0 through 5.1.2 contains an out-of-bounds read vulnerability that allows attackers to trigger a crash or read adjacent heap memory by supplying a crafted .blend file with a malicious signed short member_index value in the SDNA block. The member_index field is used as an array index into the sdna->members[] array in sdna_expand_names() without bounds validation, allowing any value outside the allocated range to produce an invalid pointer subsequently passed to strlen(), resulting in a SIGSEGV crash or unintended heap memory disclosure.

In the Linux kernel, the following vulnerability has been resolved: vsock/virtio: fix zerocopy completion for multi-skb sends When a large message is fragmented into multiple skbs, the zerocopy uarg is only allocated and attached to the last skb in the loop. Non-final skbs carry pinned user pages with no completion tracking, so the kernel has no way to notify userspace when those pages are safe to reuse. If the loop breaks early the uarg is never allocated at all, leaking pinned pages with no completion notification. Fix this by following the approach used by TCP: allocate the zerocopy uarg (if not provided by the caller) before the send loop and attach it to every skb via skb_zcopy_set(), which takes a reference per skb. Each skb's completion properly decrements the refcount, and the notification only fires after the last skb is freed. On failure, if no data was sent, the uarg is cleanly aborted via net_zcopy_put_abort(). This issue was initially discovered by sashiko while reviewing commit 1cb36e252211 ("vsock/virtio: fix MSG_ZEROCOPY pinned-pages accounting") but was pre-existing.

In the Linux kernel, the following vulnerability has been resolved: Bluetooth: hci_conn: Fix memory leak in hci_le_big_terminate() hci_le_big_terminate() allocates iso_list_data via kzalloc_obj but returns 0 without freeing it when neither pa_sync_term nor big_sync_term flags are set after evaluating the PA and BIG sync connection state. This early-return path was introduced when hci_le_big_terminate() was refactored to take struct hci_conn instead of raw u8 parameters, adding PA/BIG flag evaluation logic. The existing kfree() on hci_cmd_sync_queue failure does not cover this path.

A vulnerability was detected in CodeAstro Simple Online Leave Management System 1.0. This affects an unknown part of the file /SimpleOnlineLeave/admin/accept.php of the component POST Handler. Performing a manipulation of the argument appid results in sql injection. The attack is possible to be carried out remotely. The exploit is now public and may be used.

The Tempo Operator's gateway component failed to consistently apply namespace-scoped redaction on some query API response paths when query RBAC was enabled, allowing an authenticated user to read span attributes belonging to other tenants' namespaces.

A security vulnerability has been detected in CodeAstro Simple Online Leave Management System 1.0. Affected by this issue is some unknown functionality of the file /SimpleOnlineLeave/admin/deletemp.php. Such manipulation of the argument ID leads to sql injection. The attack can be executed remotely. The exploit has been disclosed publicly and may be used.

Mattermost versions 11.7.x <= 11.7.2, 11.6.x <= 11.6.4, 10.11.x <= 10.11.19 fail to check the manage_shared_channels permission in the /share-channel autocomplete handler, which allows an authenticated user without that permission to enumerate configured remote cluster connection metadata via slash command autocomplete.. Mattermost Advisory ID: MMSA-2026-00676

Mattermost versions 11.7.x <= 11.7.1, 11.6.x <= 11.6.4, 10.11.x <= 10.11.19 fail to restrict metric configuration changes to the playbook being saved, which allows an authenticated user with team access to alter another user’s playbook metric settings via a crafted import or update request with a foreign metric ID. Mattermost Advisory ID: MMSA-2026-00653