CVE Feed

Headroom compresses data before the data reaches a large language model. Prior to version 0.35.0, the Headroom WebSocket server does not validate the `Origin` header of incoming client WebSocket requests before forwarding the request to the upstream server, allowing malicious WebSocket clients to perform arbitrary LLM requests without authentication. This can be exploited by a malicious WebSocket client executed in a traditional or headless browser such as lightpanda, if the browser has access to the Headroom proxy and the OpenAI API key is stored in the `OPENAI_API_KEY` environment variable. Version 0.35.0 fixes the issue.

NetBSD contains an information disclosure vulnerability in mm_open() within sys/dev/mm.c that allows unprivileged local users to obtain real kernel virtual addresses by opening world-accessible devices such as /dev/null or /dev/zero, which incorrectly receive the PK_KMEM process flag. Attackers can exploit this misconfigured flag to bypass the CANSEE_KPTR obfuscation mechanism and read kernel virtual addresses for sensitive kernel structures including struct proc, kauth_cred, filedesc, and vmspace via sysctl KERN_PROC queries.

NetBSD contains a use-after-free and double-free vulnerability in msg_recv_copyin() within the COMPAT_NETBSD32 compatibility layer due to a missing return statement before the cleanup label on the success path. Any local user able to execute a 32-bit binary on a 64-bit NetBSD system can trigger a kernel panic or memory corruption by calling recvmsg() with msg_iovlen between 9 and IOV_MAX, causing the kernel to access a freed iovec buffer and subsequently free the same allocation a second time.

An information leakage vulnerability exists in the Endpoint DLP component (epdlpdrv.sys) of Netskope Client for Windows prior to version R141. An internal communication channel used by the user-space hook DLL to pass messages through the kernel driver to the daemon lacked proper token-based message validation, allowing local unprivileged processes to send unauthorized queries. Additionally, a reply buffer used by the port message handler was not properly initialized before returning data, leaking residual kernel pool memory from prior allocations. A local unprivileged attacker could exploit this vulnerability to enumerate DLP configuration and feature flags, extract live session tokens, and read kernel memory fragments from other users' operations.

Improper neutralization of argument delimiters in a command ('argument injection') vulnerability in TUBITAK BILGEM Software Technologies Research Institute Pardus Pen allows Argument Injection. This issue affects Pardus Pen: before 4.2.1.

A vulnerability in the Chef Automate API gateway and identity validation path may allow an unauthenticated actor to gain elevated access to protected Chef Automate functionality under specific conditions.

Hugo is a static site generator. From v0.161.0, Hugo executes Node tools under Node's permission model, but TailwindCSS — included in the default security.exec.allow list — requires a highly permissive configuration (--allow-addons, --allow-child-process, --allow-worker). As a result, the restrictions intended by the fix for GHSA-x597-9fr4-5857 could still be bypassed, allowing a Node tool invoked during a build to read and write files outside the project's working directory. Affected versions are those after v0.43; the issue was fixed in v0.165.0 by removing tailwindcss from the default security.exec.allow list. Users who do not use TailwindCSS, or who only build trusted sites, are not affected. As a workaround, users can define a restrictive security.exec.allow list in hugo.toml.

Hugo is a static site generator. In versions after v0.123.0 and before v0.165.0, symlinks in parent directories were not dropped during direct resource lookups, allowing path confinement to be bypassed. An attacker who can place — or who convinces a site author to place — a symlink inside a mounted directory (for example, in a locally vendored theme under themes/) can cause functions that perform direct lookups, such as resources.Get and os.ReadFile, to follow that symlink and read files outside the intended project boundaries, disclosing their contents in the built site. Themes mounted as Go modules fetched from GitHub have symlinks stripped on download and are not affected, and multi-directory walks (e.g. content/asset walking) are not affected. This issue is an incomplete-fix follow-up to GHSA-c3wq-j5vh-68rc and GHSA-fw87-fv5r-9fpw; it is fixed in v0.165.0.

AVideo through 29.0 contains an insecure direct object reference (IDOR) vulnerability in objects/categoryDeleteAssets.json.php. The endpoint validates only the Category::canCreateCategory() capability and a CSRF nonce before passing the attacker-supplied $_REQUEST['id'] to Category::deleteAssets(), which recursively removes {systemRootPath}videos/categories/assets/{id}/. It omits the Category::userCanEditCategory() ownership check enforced by the sibling Category::delete(). On installations where the non-default usersCanCreateNewCategories setting is enabled, an authenticated non-admin user with the canUpload capability can send a POST request with an arbitrary category ID and recursively delete any category's on-disk asset directory (icons/images). Category records and videos are not deleted. As of the advisory publication no patched version was available.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a stored cross-site scripting vulnerability in the Bookmark plugin where chapter names are not encoded before being concatenated into public watch-page HTML. A video owner can inject malicious scripts via the bookmark name parameter, and every visitor of that video executes the payload in the AVideo origin.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a stored cross-site scripting vulnerability in the LoginControl plugin that fails to HTML-encode PGP public keys echoed into a textarea element. An authenticated attacker can inject malicious JavaScript by submitting a crafted public key, which executes in an administrator's session when viewing the user's profile tab.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a stored cross-site scripting vulnerability in the CustomizeUser plugin where the field_name parameter is stored raw without sanitization. Administrators can inject malicious scripts via the add.json.php endpoint that execute when viewing extra info pages or profile forms that render the typeToHTML function.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a stored cross-site scripting vulnerability in the user 'donationLink' profile field. User::setDonationLink() (objects/user.php) stores the value and save() validates it only with filter_var(..., FILTER_VALIDATE_URL), which accepts strings such as http://evil.example/"onmouseover=alert(document.domain)//, while getDonationLink() applies only strip_tags() and does not encode double quotes. plugin/CustomizeUser/actionButton.php echoes the value unencoded into an <a href="..."> attribute, and that button is included from view/modeYoutubeBottom.php on the watch page when the CustomizeUser option allowDonationLink is enabled. An authenticated user who updates their own profile via objects/userUpdate.json.php can therefore break out of the href attribute and inject an event handler that executes JavaScript in the browser of any visitor—including an administrator—who views the attacker's videos and interacts with (for example, hovers over) the donation button. The issue was unfixed at the time of reporting.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 fails to verify ownership in addLiveLink.php when updating LiveLinks, allowing authenticated users to modify other users' links. A canStream user can overwrite another user's LiveLink HLS source and metadata by supplying an existing linkId, redirecting viewers to attacker-controlled media.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 fails to validate ad impressions in plugin/AD_Server/log.php, allowing logged-in users to submit arbitrary label values that trigger unverified wallet credits to campaign video owners. Attackers can repeatedly POST label=start requests to mint YPTWallet balance for any campaign video without proof an ad actually played.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains an unauthenticated file read vulnerability in the getRecordedFile.php endpoint that streams recorded FLV files from the temporary directory. Attackers can request the endpoint with a known or guessed stream key to download recorded live video files without authentication or authorization checks.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a stored cross-site scripting vulnerability in the YPTWallet plugin where user-supplied CryptoWallet values are base64-encoded but not HTML-escaped before storage in wallet_log.information. Administrators viewing pending withdrawal requests in pendingRequests.php execute the stored markup in their session, allowing attackers to perform administrative actions via same-origin fetch requests.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 is missing an authentication/authorization check in plugin/WebRTC/status.json.php. When the WebRTC plugin is present, any unauthenticated remote user can request /plugin/WebRTC/status.json.php and receive JSON containing the absolute filesystem path of the WebRTC2RTMP helper binary (revealing the document-root path), the configured WebRTC port, file_exists/is_executable status for the binary, the contents of the WebRTC log/JSON files (videos/WebRTC2RTMP.log) when present, and whether the configured port is reachable on loopback (127.0.0.1) and on the public address. The endpoint performs no User::isLogged(), User::isAdmin(), or forbiddenPage() check. The issue was unfixed at the time of reporting.

WWBN AVideo at commit c3edcc274c389816d434acadac07ee78eaf330c1 and earlier contains an XML injection vulnerability in plugin/AD_Server/VMAP.php, which is reachable without authentication when the AD_Server plugin is enabled. The script emits Content-Type: application/xml and writes the timeOffset and idTag values returned by AD_Server::getVMAPSFromRequest() directly into VMAP attributes without encoding. Those values originate from the base64- and JSON-decoded $_REQUEST['vmaps'] parameter, which is not covered by $securityFilter in objects/security.php. A remote attacker can craft a vmaps value (containing a non-empty VAST.campaing entry) and induce a user to open the resulting VMAP URL or a video page using it, injecting arbitrary <vmap:AdBreak> and <vmap:AdTagURI>/AdSource nodes into the generated XML. The player's IMA path (afterVideoJS -> PlayerSkins::setIMAADTag) then requests the attacker-supplied ad URLs, resulting in ad injection and cross-origin requests from the victim's playback session. The issue was unfixed at the time of reporting.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a CSV formula injection vulnerability in the myComments.download.php endpoint that fails to sanitize spreadsheet formula prefixes in comment text. Authenticated users can inject formulas starting with =, +, -, or @ characters that execute when administrators or video owners open the exported CSV file in spreadsheet applications.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a cross-site request forgery vulnerability in playlistRemove.php that allows attackers to delete playlists by skipping CSRF protection checks. Attackers can craft a malicious form that submits a POST request to playlistRemove.php, causing a victim's playlist to be deleted when they visit the attacker's page while logged in.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a reflected cross-site scripting vulnerability in plugin/Gallery/view/Category.php when SubCategorys is enabled. The getBackURL parameter is echoed into an href attribute without HTML encoding, allowing attackers to inject malicious scripts that execute in visitors' browsers when they interact with the Back control.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a stored cross-site scripting vulnerability in UserGroups::setGroup_name() that fails to sanitize group_name input. Administrators with canAdminUserGroups permission can inject malicious HTML and JavaScript that executes in the browser when other administrators access the user manager interface.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a server-side request forgery vulnerability in the _json_decode function that fetches remote URLs and local file paths without SSRF validation. Unauthenticated attackers can POST file paths or HTTP URLs to login.json.php to read local files or access internal services, with results parsed as login credentials.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a reflected cross-site scripting vulnerability in confirmLivePassword.php that copies REQUEST_URI into a form action attribute without encoding. Attackers can craft a malicious URL with a quote character to break out of the action attribute and inject event handlers that execute in the victim's browser within the site origin.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a reflected cross-site scripting vulnerability in plugin/Live/confirmLivePassword.php. The script interpolates the unauthenticated GET parameter u (which is not covered by $securityFilter) directly into an <img src="..."> attribute without URL- or HTML-encoding. A remote attacker can craft a link containing a double-quote character in u (with a non-empty key parameter and no valid c parameter) to close the src attribute and inject an additional tag with an onerror handler, executing arbitrary JavaScript in the site's origin in the browser of any user, including an administrator, who opens the link. No patched version was available at the time of the advisory.

WWBN AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains a reflected cross-site scripting vulnerability in the showAlertMessage() function that inserts the raw Referer header into a JavaScript comment without encoding. Attackers can craft a Referer header containing */ to close the comment and inject arbitrary JavaScript that executes in the site origin for visitors arriving from the attacker-controlled page.

AVideo through commit c3edcc274c389816d434acadac07ee78eaf330c1 contains an open redirect in objects/playlistSort.php. Because the endpoint is not a *.json.php script, AVideo's automatic CSRF guard (autoCSRFGuard()/forbidIfIsUntrustedRequest()) does not run, and when the request includes the sort parameter the script issues a Location header set to the unvalidated $_SERVER['HTTP_REFERER'] value without calling isSafeRedirectURL(). A remote unauthenticated attacker can therefore induce a logged-in user who can manage the targeted playlist to submit a cross-origin POST with a crafted Referer, causing the victim's playlist to be reordered and the victim's browser to be redirected from a trusted AVideo URL to an attacker-controlled site for phishing. No patched version is available.

compression is a Node.js and Express compression middleware. In versions before 1.8.2, when a client aborts the connection while a compressed response is still being sent, the zlib stream created to compress that response is never destroyed, so each aborted compressed response leaks its native zlib memory. A remote unauthenticated attacker can repeatedly open requests and disconnect early, exhausting the available memory and crashing the server. All applications using compression are affected. The issue is fixed in compression 1.8.2, and users should upgrade to 1.8.2 or later.

SGLang allows unauthenticated pickle deserialization through /update_weights_from_tensor when no auth keys are configured, and the SafeUnpickler policy can be bypassed because builtins.import and builtins.getattr are resolvable, enabling code execution via pickle REDUCE.

The BuddyPress plugin for WordPress is vulnerable to Insecure Direct Object Reference in all versions up to, and including, 14.3.3 via the bp_notifications_action_bulk_manage due to missing validation on a user controlled key. This makes it possible for authenticated attackers, with Subscriber-level access and above, to delete/mark as read/mark as unread notifications of other users.

Net-SNMP through 5.9.5.2 contains a denial of service vulnerability in the SMUX module where smux_accept() performs an unauthenticated blocking read without timeout on newly accepted connections. An unauthenticated remote client can connect to the SMUX listener and send no data, causing the single-threaded snmpd main loop to block indefinitely and suspend all SNMP processing.

libp2p-rendezvous through 0.17.1 fails to validate registration TTL values in discovery responses, allowing attackers to trigger timer arithmetic overflow. A malicious rendezvous server can send a discovery response with an unbounded TTL value that causes the client node process to panic when computing the expiry timer.

The MetForm WordPress plugin before 4.1.9 does not properly neutralize newline characters in user-submitted values that are placed into notification email headers, allowing unauthenticated attackers to inject additional email headers, such as Bcc, into the emails the site sends when a submitted field value is configured to populate a header.

The Persian Elementor WordPress plugin from 2.7.10 before 2.8.2 does not verify that the payment authority returned to its ZarinPal payment callback belongs to the transaction being completed, allowing unauthenticated attackers to complete a pending order using a valid payment authority obtained from a different transaction.

The Simple CAPTCHA with Cloudflare Turnstile WordPress plugin from 1.2.2 before 1.42.3 runs the shortcode parser over the whole rendered Contact Form 7 form, including the values a visitor submitted, allowing unauthenticated users to execute arbitrary shortcodes registered on the site.

The Payment Gateway PayPay for WooCommerce WordPress plugin from 0.5 to 0.9.3 does not verify the authenticity of the payment notifications it receives before acting on them, allowing unauthenticated attackers who know the store's merchant identifier to mark arbitrary orders as paid, or to cancel or fail them.

The Nexi XPay Build WordPress plugin from 7.6.1 to 7.6.2 does not verify that the saved payment token being requested belongs to the current user, allowing unauthenticated attackers to retrieve other customers' stored card token references together with a valid authorisation signature.

A symlink-following flaw was found in libvirt's qemuTPMEmulatorPrepareHost() function. The function uses a path-based chown() on the swtpm logfile without checking for symbolic links. A local attacker with access to the swtpm account can replace the logfile with a symlink, causing libvirtd (running as root) to transfer ownership of an arbitrary file to the swtpm user.