Server-side request forgery in phpMyFAQ's PDF export feature lets an attacker who can plant an image URL inside FAQ answer content make the server fetch a destination of their choosing - for example a loopback or internal HTTP service - and embed the response in the generated PDF. All releases before 4.1.10 are affected, and the vendor has released 4.1.10 as the fix. The reported chain requires two preconditions: permission to create or modify the FAQ containing the malicious <img> reference, and the target host being permitted by records.allowedMediaHosts (the reporter explicitly added 'localhost' to that allowlist, a non-default setting); the reporter could not demonstrate the full chain against a completely unauthenticated attacker or against default configurations, and confirmed only that retrieval and embedding occur, not that arbitrary internal data can be extracted. No CISA KEV entry or EPSS score was supplied, and the advisory ships a local reproduction script rather than verified working exploit code; the priority signal is a routine monthly patch cadence.
Cross-site request forgery in LinkStack (linkstackorg:linkstack) affects all versions up to and including 4.8.8 and lets a remote, unauthenticated attacker cause a logged-in user's browser to submit state-changing requests against the instance. Exploitation depends on an authenticated victim with an active session being tricked into interacting with attacker-controlled content, so a successful attack yields only a low integrity impact on LinkStack itself, with no confidentiality or availability effect. No vendor patch or advisory has been issued - the maintainers were contacted about the disclosure and did not respond - and no public exploit code or proof-of-concept has been identified at time of analysis; the reported CVSS 4.0 score of 5.3 and the this_month priority label place this as a moderate, non-urgent item to schedule rather than an emergency.
Stored user credentials in SH3, within the Módulo de Gestão and Módulo de Arrecadação components, are protected only by a reversible XOR-based encoding rather than a computationally expensive one-way password hash (CWE-916), so anyone able to read that credential data can reverse it back to plaintext. Only authenticated, high-privileged users (CVSS PR:H) can reach the affected data, and the vendor rates exploitability as difficult with high attack complexity; impact is limited to confidentiality of stored credentials with no integrity or availability effect, giving a low severity score (CVSS 4.0 base 2.1, independently assessed as CVSS 3.1 AV:N/AC:H/PR:H/UI:N/S:U/C:L/I:N/A:N). A public PoC reference exists in the form of a Medium write-up; it is not verified as working exploit code, and there is no KEV listing or other indication of active exploitation. Priority is monitor.
Stored cross-site scripting in WWBN AVideo 26.0 through 29.3.0 allows any user who holds comment permission on a viewable video to persist JavaScript that executes in other viewers' browsers under the application origin. The payload must begin an HTML block with an image tag carrying an event handler, because the custom Parsedown subclass sanitizes inline image/link markup but hands leading block-level `<img>`/`<a>` tags to the raw HTML handler. Exploitation requires an authenticated commenter, comment visibility, and a victim viewing the comment (UI:P); impact is bounded to low confidentiality/integrity per the assessed vector, and a CSP that blocks inline event handlers neutralizes the demonstrated payload. No public exploit code or confirmed active exploitation was identified, and no verified patched release exists as of the advisory date.
Stored cross-site scripting in YesWiki before 4.7.0 lets any authenticated low-privilege user upload an unsanitized SVG through the bazar ImageField, which the server stores byte-identical and serves as image/svg+xml, so script executes in the browser of anyone who opens the entry or navigates to the stored file. The vulnerable sink is a sibling of the one fixed in 4.6.7 for GHSA-67v5-7pfv-pp9m (CVE-2026-104461): the cleanFile() denylist was added only to FileField, leaving ImageField with an isImage() gate that passes svg because it sits in the default attach_config.ext_images list, so stock configurations are exposed. A vendor fix is available in 4.7.0, and the GitHub advisory documents a public proof-of-concept with recorded live reproduction on v4.6.7; it is not verified here as working exploit code, and there is no CISA KEV signal of active exploitation.
Stored cross-site scripting in SiYuan Note 2.10.0 through 3.8.6 lets a user who already holds edit permission on a document plant a `javascript:` URL inside math-block content, for example `$$\href{javascript:alert('xss')}{clickme}$$`, so that script runs in the application origin when a reader clicks the rendered link. The rendering path passes math TeX to KaTeX with `trust: true` and mounts the resulting HTML through `innerHTML`, so `\href`, `\url` and `\includegraphics` payloads (including mixed-case `JaVaScRiPt:` schemes) survive unsanitized. Exploitation is not unauthenticated and is not automatic: the attacker needs writer access to a shared document and the victim must open it and click, which caps practical severity at CVSS 5.1; the vendor advisory GHSA-hhvq-jqv9-pgjq includes a detailed proof-of-concept reproduction against a real Chromium instance, though no CISA KEV listing or public weaponized exploit is reported, and the priority label is this_month.
Path traversal in the EventCatalog (@eventcatalog/sdk) custom-documentation helpers, fixed in 2.30.0, lets a caller that already has local access to the SDK escape the catalog docs directory when reading, writing, or deleting custom docs. The vulnerable functions getCustomDoc, getCustomDocs, writeCustomDoc and rmCustomDoc joined attacker-supplied path values directly onto the docs directory, so '../' sequences, absolute or leading-slash paths, Windows drive-letter or UNC paths, null bytes, or symlinks could redirect file operations outside the catalog. Exploitation is limited to local-host callers who can already invoke the SDK - typically local build or documentation tooling rather than a network service - with low confidentiality, integrity and availability impact; no public exploit code was identified at time of analysis and the issue is rated moderate with low urgency.
Stack exhaustion in ImageMagick's MVG decoder can crash the process, causing a denial of service limited to availability; no code execution, data disclosure, or integrity impact is assessed. ImageMagick 6.x before 6.9.13-58 and 7.x before 7.1.2-33 are affected, and a crafted MVG file must be processed through certain decoder paths with specific arguments, requiring local access and user interaction or an automated pipeline. Vendor patches are available in 6.9.13-58 and 7.1.2-33, and no public exploit has been identified at time of analysis.
ImageMagick's PS, PDF and XPS decoders allow a security-policy limit on scene selection to be bypassed by supplying a specific argument, so a crafted input can make the tool process more scenes than the configured policy permits and consume resources. Only installations that define such a list-length limit in their security policy have a limit to circumvent; the CVSS 4.0 vector (AV:L/AC:L/PR:L/UI:N) scores it 4.8 and confines impact to low availability, with no confidentiality or integrity effect. Fixes are released in 6.9.13-58 and 7.1.2-33, no public exploit code or PoC reference is identified in the supplied sources, and the issue is not flagged as actively exploited (no CISA KEV entry); priority is 'this_month' rather than emergency.
ImageMagick 6.x before 6.9.13-58 and 7.x before 7.1.2-33 can be driven into memory exhaustion when the -cdl (color decision list) option processes attacker-influenced input, because allocated memory is not released along that code path. The issue is a CWE-401 memory leak scored 4.8, with local reachability and passive user interaction in the assessed CVSS 3.1 vector (AV:L/AC:L/PR:N/UI:R, availability-only Low), and exploitation depends on the -cdl option being reachable in the deployment. No public exploit was identified at time of analysis; vendor fixes are available in 7.1.2-33 and 6.9.13-58.
PerlMagick, the Perl binding shipped with ImageMagick, leaks memory in five property-reading code locations before release 7.1.2-33, so an attacker who can repeatedly drive property reads through that binding can exhaust process memory and cause a denial of service. Exposure is confined to applications or scripts that actually use PerlMagick and read image properties - the core C library and other language bindings are not affected through this path, and a single read leaks only a trivial amount, so the reads must recur. The vendor has released a fix in 7.1.2-33; no public exploit code was identified at time of analysis and the issue is not in CISA KEV, which fits its genuinely low priority despite the moderate 4.8 CVSS 4.0 score.
A missing memory release in ImageMagick's `-set profile` handling causes each invocation that touches that code path to leak memory, which can degrade or exhaust resources on hosts that repeatedly process attacker-influenced images. The flaw affects 6.x builds before 6.9.13-58 and 7.x builds before 7.1.2-33; exploitation requires a user or automated job to be induced to process a crafted image with the `-set profile` option, giving a local vector with passive user interaction and no privileges required. Impact is confined to availability - a single run leaks rather than crashes, so a meaningful denial of service needs repeated processing - and no public exploit was identified at time of analysis.
Heap buffer overflow in ImageMagick's X11 animate path (versions before 6.9.13-58 and 7.x before 7.1.2-33) allows an attacker-supplied crafted image to overwrite heap memory and crash the application when a user runs an X11 animate operation with an X11 display available. Exploitation requires local user interaction and the available assessment limits impact to availability; no public exploit was identified at time of analysis. Vendor patches are available in 6.9.13-58 and 7.1.2-33.
MVG encoder injection in ImageMagick before 6.9.13-58 and 7.x before 7.1.2-33 lets an attacker who can supply drawing options write arbitrary MVG drawing commands into generated MVG output. The flaw is medium severity (CVSS 4.0 base 4.8; assessed CVSS 3.1 vector AV:L/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N), requires a local attack path and a downstream step that processes the resulting MVG file, and affects applications that pass untrusted drawing options to ImageMagick. No public exploit identified at time of analysis; vendor fixes are available in 6.9.13-58 and 7.1.2-33.
Heap buffer over-read in ImageMagick before 7.1.2-33 can disclose adjacent heap memory or crash the process when a crafted image is processed through the minimum-bounding-box printing path. The issue is local, requires no privileges, and depends on a victim or automated service feeding attacker-controlled image data into that specific operation, with impact limited to read-only disclosure and availability loss. No public exploit was identified at time of analysis, and the vendor has released 7.1.2-33 as the fixed version; the priority signal is this_month.
ImageMagick 6.x before 6.9.13-58 and 7.x before 7.1.2-33 can run an executable that policy.xml is intended to block, because delegate invocations are parsed incorrectly. The bypass matters only where an administrator has configured policy.xml to deny a delegate or executable and where crafted input reaches delegate processing; the assessed vector is local (AV:L) with user interaction (UI:R) and integrity-only impact, and no public exploit was identified at time of analysis. Vendor patches are available in 6.9.13-58 and 7.1.2-33.
Memory exhaustion in ImageMagick before 7.1.2-33 stems from a leak in the FPX (FlashPix) encoder, so unauthenticated remote attackers who can push crafted images through that encode path can consume process memory incrementally and degrade availability. Exposure is narrow: the leak is confined to FPX output encoding rather than general image handling, and denial of service depends on accumulating many small leaks across repeated requests, which is why the vendor vector (CVSS 4.0, 2.3) records high attack complexity, a present attack requirement and availability-only impact. No public exploit code or confirmed active exploitation was identified at time of analysis, and the vendor has released 7.1.2-33 as the fix.
OpenRemote Core before 1.31.0 allows a registered edge gateway to overwrite attributes on, or delete, assets in any realm, breaking tenant isolation. The gateway event path skips normal authorization and relies on a reversible GatewayConnector.mapAssetId transform, so an attacker with one registered, enabled gateway's service-user credentials can compute pre-image asset IDs that resolve to victim assets across tenants. Assessment signals conflict sharply - the supplied CVSS 4.0 vector scores 2.3, while the independent CVSS 3.1 assessment rates it 9.6 (critical) due to scope change and high integrity/availability impact; it is not in CISA KEV, and a public PoC reference exists but is not verified as working exploit code.
OpenRemote Manager before 1.31.0 routes incoming gateway WebSocket traffic by gateway ID alone, so a user holding WRITE:ADMIN in one realm can create a service user named after a gateway in another realm and impersonate that gateway across realm boundaries. The assessed impact is limited destructive tampering (third-party descendant assets can be created or overwritten, but existing assets cannot be read) plus the ability to establish SSH tunnels into the target gateway's local network during the initial sync handshake. This is an authenticated, high-privilege attack (CVSS 3.1 PR:H, AC:H) that additionally requires knowledge of the victim realm's exact gateway ID; no CISA KEV entry or confirmed working exploit code is available, though the vendor advisory documents a step-by-step proof-of-concept procedure, and the current decision signal is monitor rather than emergency action.
An authenticated low-privileged user of KodCloud KodExplorer up to and including version 4.55 can write arbitrary files through the Online Editor save endpoint in app/controller/editor.class.php by manipulating the path and filestr arguments, yielding an unrestricted upload restricted in scope to that user's accessible storage. Impact is limited (low confidentiality, integrity and availability) per both the CVSS 4.0 vector (5.3) and an independent CVSS 3.1 assessment of AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L; remote exploitation requires a valid account and no user interaction. The flaw was disclosed publicly with exploit details, and the vendor did not respond to early contact, so no vendor-released patch has been identified at time of analysis.
Edgeless Systems Contrast releases 1.21.0 through 1.24.1 can apply a wildcard registry pull configuration to a concrete registry domain in the imagepuller guest component, potentially exposing registry credentials, CA bundles, or mirror/pull metadata to the wrong audience. The bug only matters in deployments whose pull configuration lists both a concrete domain and its parent domain as a wildcard AND where the wildcard's credentials, CA bundle, or mirror settings are unsuitable for that concrete registry - the vendor describes such overlapping setups as very unlikely in practice. Exploitation requires low privileges and adjacent network positioning, severity is low (CVSS 4.0 base 2.1; assessed CVSS 3.1 AV:A/AC:H/PR:L/UI:N/S:U/C:L/I:N/A:N), and no public exploit code has been identified; the issue is fixed in 1.25.0 and warrants routine monitoring rather than urgent action.
SMS multi-factor authentication login-code requests in Parse Server are persisted with master privileges, letting an authenticated user who is enrolled in SMS MFA smuggle extra account fields into that request so they are written to their own record with restrictions bypassed. Successful abuse allows integrity-only changes to self-owned data, such as flipping the email-verification flag or setting values normally protected by beforeSave triggers; other users' records cannot be touched. The flaw affects only deployments using the built-in SMS MFA adapter with enrolled accounts, in versions before 8.6.104 and 9.0.0 through 9.10.5-alpha.5, and no public exploit identified at time of analysis while vendor patches are available.
SQL injection in pig-mesh pig through 4.1.0 lets an authenticated user inject SQL into ORDER BY clauses by supplying crafted `ascs`/`descs` sort parameters to endpoints that bind pagination through SqlFilterArgumentResolver. The pre-patch code relied on MyBatis-Plus SqlInjectionUtils.check, which the upstream fix's own tests show fails to block case/when expressions, hex(), arithmetic such as user_id+1, comment sequences, and backtick-quoted identifiers. A public PoC reference exists (a GitHub write-up); it is not verified as working exploit code, there is no confirmed active exploitation, and the vendor-released fix is version 4.1.1 (commit 9b6311161e346d7626c81d993044072fed2dca07).
OS command injection in dbcli mycli through 2.30.1 lets a local user who can drive the \source special command slip past the source_special_command_is_safe whitelist check and execute commands with that user's own privileges. Affected builds are mycli 2.29.0 through 2.29.4 and 2.30.0 through 2.30.1, and the disclosure states that exploit code has been made publicly available, with the CVSS 4.0 vector scoring E:P for a public proof-of-concept. Impact is bounded because the attack is local only (AV:L) and runs inside the invoking user's context (VC:L/VI:L/VA:L), making this a low-to-moderate priority issue despite the OS command injection class; no CISA KEV listing is present in the supplied data.
Repeated failed comparisons in ImageMagick's `magick compare` command (every release before 7.1.2-33) leak memory that is never released, allowing local attackers to gradually exhaust the memory of a process that runs compare against attacker-influenced images and thereby degrade availability. Each failure leaks only a small amount, so sustained or repeated failed invocations are needed before service quality drops; there is no data exposure or code execution impact. No public exploit code was identified at time of analysis, the issue is not in CISA KEV, and the vendor priority label is monitor.
Heap memory corruption in ImageMagick versions before 7.1.2-33 affects only 32-bit builds and can crash the process when a crafted image is processed by a workflow that requests convex-hull information. The assessed impact is availability only; code execution is described as potential and is not confirmed. Exploitation requires the 32-bit build plus the convex-hull path, no public exploit has been identified at time of analysis, and the vendor has released 7.1.2-33.
Cross-site scripting (CWE-79) in the Onyx chat/knowledge platform through version 5.0.0, specifically in the Craft feature's output panel rendered by web/src/app/craft/components/output-panel/PreviewTab.tsx, lets an attacker inject script that executes in the browser context of another user who views the crafted preview. Exploitation requires a low-privileged authenticated account on the Onyx instance plus a victim who renders the malicious preview, and the vendor-assessed impact is limited to integrity (no confidentiality or availability loss), which is why VulDB scored it 5.1; a public exploit disclosure accompanies the report, but it is not independently verified as working exploit code, and there is no KEV listing or evidence of active exploitation. Onyx administrators should treat this as a moderate-priority hardening item to be scheduled within the current month rather than an emergency.
ImageMagick before 6.9.13-58 and 7.x before 7.1.2-32 can bypass configured security-policy restrictions because the fix for GHSA-jjp4-3fwf-393j was incomplete. The exposure is limited to deployments that actually rely on policy.xml to block specific coders or operations and that accept attacker-supplied input; the vendor CVSS 4.0 vector is local (AV:L), high complexity (AC:H), with a present attack requirement (AT:P) and high privileges (PR:H), and low impact across confidentiality, integrity and availability. No public exploit identified at time of analysis; CISA KEV does not list it and the vuln.today decision is monitor.
Starman versions before 0.4019 for Perl allow memory exhaustion via unbounded buffering of chunked request bodies. The chunked body decoder keeps each chunk in memory until the whole chunk and its CRLF have arrived, and sets no limit on the length of a chunk-size line. A client that declares a huge chunk size and streams data, or sends a chunk-size line that never ends, grows the buffer for as long as it keeps sending. A remote attacker can exhaust worker memory with a single connection.
Starman versions before 0.4019 for Perl send response bodies for HEAD requests, desynchronizing responses on keep-alive connections. For a HEAD request Starman omits chunked framing but still writes the body returned by the application, on both the array and streaming response paths. Responses with a 1xx, 204 or 304 status also carry the body. On a keep-alive or pipelined connection the client reads those bytes as the start of the next response. A front-end proxy that reuses upstream connections may then pair responses with the wrong requests.
Starman versions before 0.4019 for Perl allow CPU exhaustion via quadratic regex backtracking when parsing chunked request bodies. Starman decodes chunked request bodies before calling the application, matching each chunk-size line with `s/^(([0-9a-fA-F]+).*\015\012)//`. A line of hex digits followed by a bare LF and then CRLF never matches, because `.` does not match LF, and each failed attempt backtracks in time quadratic in the length of the hex run. Nothing limits the length of the line. A remote attacker can exhaust worker CPU by sending such requests to any route.
Starman versions before 0.4019 for Perl allow a stalled request body to hold a worker indefinitely, as read_timeout covers only the headers. Content-Length and chunked request bodies are read with blocking reads and no deadline. A client that sends complete headers and then stalls or trickles the body keeps the worker waiting until it disconnects. A remote attacker can take every worker with a few such connections, five with the default worker count, so the server stops answering requests. Version 0.4019 applies read_timeout to each wait for body data and adds a body_timeout limit, 300 seconds by default, on the total time to read a body.