Stored cross-site scripting in EspoCRM versions prior to 9.3.9 allows an authenticated user holding Email Template or PDF Template edit permission to persist crafted HTML that evades the moderateSanitizeHtml() filtering routine. The payload only executes when a second user opens the tampered template in edit mode on an instance where the non-default setting clientCspDisabled is true, at which point the attacker can hijack the victim's session; default deployments keep Content-Security-Policy enabled and are not affected. Publicly available exploit code exists, but there is no confirmed active exploitation (CISA KEV) and the vendor has shipped a fix in 9.3.9, making this a genuine but low-priority, non-default-configuration-dependent issue.
OS command injection in Netgate pfSense CE 2.5.0 through 2.9.0 allows an authenticated OpenVPN client to run arbitrary shell commands as the low-privileged 'unbound' user by embedding shell metacharacters in the certificate common name or username that openvpn.learn-address.sh interpolates unquoted into a 'su -c' string. Exploitation requires three conditions to line up: valid VPN authentication (the CVSS vector is PR:L, so authenticated), the non-default 'DNS Resolver (unbound) client registration' feature enabled, and attacker-controlled CN or username - which is why the assessed severity is low (CVSS 2.3 in the v4 vector supplied; equivalent CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:L per independent assessment). Publicly available exploit code exists; there is no confirmation of active exploitation (not in CISA KEV) and impact is confined to the unbound user rather than root, so this is a genuine but low-priority issue.
Stored cross-site scripting in EspoCRM (all versions prior to 9.3.9) lets an authenticated administrator inject JavaScript that executes in every user's browser session by placing crafted content in the applicationName setting and abusing CSP nonce injection. The payload is persisted in the main application page without escaping, so it executes for all users who load the application and can leak page data including unsaved form input. Publicly available exploit code exists, but exploitation requires the highest application privilege (administrator), and impact is limited to confidentiality (C:L) with no integrity or availability effect, making this a genuinely low-priority issue despite the available POC.
Reflected cross-site scripting in Netgate pfSense CE 2.7.1 through 2.9.0 allows unauthenticated remote attackers to inject script into the WebGUI via the unencoded 'if' parameter in firewall_rules.php, but only when the non-default 'Require Firewall Interface' option is enabled and a privileged operator is lured into opening a crafted GET link. Publicly available exploit code exists, yet the vulnerability is a low-priority issue because exploitation requires a non-default configuration, user interaction, and a victim who already holds firewall-rules privileges, with impact confined to that victim's session.
An out-of-bounds read in ggml-org whisper.cpp through version 1.9.4 lets a local attacker crash the application by getting a victim to load a crafted GGML model file whose tensor-type field ('ttype') falls outside the valid range. Exploitation is strictly local and requires user interaction - the victim must choose to open the untrusted model - and the assessed impact is limited to availability (a process crash), with no confidentiality or integrity effect. Publicly available exploit code exists (the upstream issue and PR are public), but the vulnerability is not in CISA KEV and its severity is low.
Persistent cross-site scripting in the HeliumEdu Platform student planner (all versions prior to 2.2.60) lets an authenticated user upload SVG or HTML files through POST /planner/attachments/ and have embedded script execute when a victim opens the resulting public attachment URL, enabling spoofed content and phishing under an application-managed storage URL. Exploitation is constrained: the attacker must already hold an account permitted to use the attachments endpoint, and the victim must actively open the link (CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:U/C:N/I:L/A:N, base score 3.5), because attachments are served from a separate origin from the application. That origin separation means the browser's same-origin policy blocks the injected script from reading application cookies, tokens, or storage, so session or account compromise is not supported by the maintained deployment configuration; no public exploit code was identified at time of analysis and the issue is not listed in CISA KEV. This is a genuinely low-priority issue despite being network-reachable, and it is fixed in version 2.2.60.
Improper input validation in HPE NonStop KMS Server (Key Management Services) can let a locally authenticated user, with required operator interaction, trigger a buffer overflow via a crafted input message, causing a limited availability impact to the KMS Server. The assessed CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:R/S:U/C:N/I:N/A:L (score 2.8), so exploitation requires local access, low-level privileges, and user interaction, with no confidentiality or integrity effect. No public exploit code or CISA KEV active-exploitation signal was identified at time of analysis.
Authenticated buffer overflow in QNAP Qsync Central (the QsyncServer app on QNAP NAS devices) prior to 5.0.0.4 allows a remote attacker who already holds a valid Qsync Central user account to corrupt memory or crash the Qsync process, degrading availability with no confirmed path to data theft or modification. The vulnerability is a classic CWE-120 buffer copy without input-size checking, and the assessed CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/C:N/I:N/A:L) confirms the impact is availability-only and gated behind account possession. There is no public exploit identified at time of analysis, no evidence of active exploitation, and an official fix is available in Qsync Central 5.0.0.4 and later, making this a low-priority, readily remediated issue.
Denial of service in QNAP Qsync Central before 5.0.0.4 can be triggered by an authenticated low-privileged user who reaches a NULL pointer dereference in the service. The vendor has fixed the issue in Qsync Central 5.0.0.4 and later, no public exploit identified at time of analysis, and the independent assessment rates this as a genuinely low-priority, availability-only issue.
NULL pointer dereference in QNAP Qsync Central allows an authenticated remote user to cause a denial-of-service (DoS) condition that crashes the Qsync Central service. Exploitation requires the attacker to first obtain a valid Qsync Central / NAS user account, and the resulting crash is recoverable, affecting only the Qsync Central service rather than the entire NAS. No public exploit code has been identified at the time of analysis; the vendor has released a fix in Qsync Central 5.0.0.4 and later.
An authorization bypass in La Suite Docs through 6.0.0 allows a user with COMMENTER-level access to any document they control to insert comments into a thread belonging to a different, inaccessible document by naming their own document in the request path and supplying the victim thread's UUID. The attack requires prior knowledge of a valid thread UUID, which is not enumerable and must be leaked or observed, so although this is a genuine IDOR (CWE-639) the practical impact is limited to comment injection with no read access to the victim document's contents. No public exploit code or confirmed active exploitation has been identified at time of analysis, and the vendor has released a fix in versions after 6.0.0 (GitHub PR #2838).
An insecure direct object reference (IDOR) in crmeb_java through version 3.1 lets any authenticated member of a storefront read other members' order details by requesting /api/front/order/express/{orderId} or /api/front/order/apply/refund/{orderId} with someone else's order identifier, exposing tracking numbers, logistics traces, payment amounts and product lines. Exploitation is not possible pre-authentication: a valid member account (CVSS PR:L) plus a valid or brute-forceable victim order ID are both required, and the flaw is read-only with no write, modification or availability impact. No public exploit code or confirmed active exploitation has been identified at time of analysis, and the independent assessment rates this a genuine but low-severity authorization flaw rather than an emergency, reflected in a CVSS 3.1 base score of 2.3 (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N).
Authenticated cross-tenant access control failure in Papermark through 0.22.0 lets any user with a valid account delete another team's data room viewer groups, read those groups' share links, and rewrite their document permissions by issuing requests to the /api/teams/[teamId]/datarooms/[id]/groups/[groupId] endpoints with another team's identifiers. No public exploit identified at time of analysis and the flaw is not confirmed actively exploited (no CISA KEV entry), but exploitation only requires a low-privileged authenticated account plus knowledge or enumeration of a target team's teamId, data room id, and groupId, which is why the CVSS 4.0 base of 2.3 (driven by the AT:P attack-requirement metric) sits well below our independent CVSS 3.1 assessment of AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L. Impact is bounded to viewer-group deletion, share-link disclosure, and permission rewrites within multi-tenant deployments that use data room viewer groups and share links - not full account, host, or data-plane compromise - but the integrity consequences for a victim's document-sharing posture are real, and the exposed share links should be treated as disclosed.
Improper access control in Lychee before 7.6.1 allows remote attackers to probe for and alter the highlight (star) state of photos they are not permitted to see, including photos in private or password-protected albums. The flaw only exists on installations configured with the non-default setting photos_star_visibility set to 'anonymous', where PhotoPolicy::canHighlight() returns true for callers without performing a canSee check, and SetPhotosHighlightedRequest::processValidatedValues() runs before authorization. Impact is narrow - existence disclosure via differing success/not-found responses plus toggling of is_highlighted - with no read of photo content, no deletion and no availability impact; exploitation additionally requires guessing valid photo identifiers, and no public exploit code has been identified at time of analysis (no CISA KEV listing). Note the source description labels the attacker 'unauthenticated' while the supplied CVSS 4.0 vector carries PR:L, and our independent assessment scores it PR:N, so authentication requirements are not fully confirmed from available data and should be verified with the vendor.
Missing record-level authorization in metasfresh through 5.175 allows an authenticated low-privilege user to bypass role and organization access checks by calling the BoardRestController handlers getBoard, getCard, addCard, removeCard and patchCard, which lets them read records outside their scope and modify board card assignments. The CVSS 4.0 vector confirms the prerequisites are an authenticated attacker (PR:L) over the network with no user interaction, and the independent CVSS 3.1 assessment (low confidentiality and integrity impact, scope unchanged) agrees that exploitation is limited by the need for valid credentials plus knowledge or enumeration of out-of-scope record identifiers. No public exploit code is identified at time of analysis and CISA KEV does not list this CVE, so both the reported 2.3 score and the independent assessment characterize this as a genuinely low-priority, confined authorization weakness rather than a high-impact compromise vector.
Server-side request forgery in the Lychee photo-management tool before 7.6.0 lets an authenticated user who can reach the POST /api/v2/Photo::fromUrl URL-import endpoint steer server-side fetches toward private or otherwise forbidden destinations by manipulating DNS answers between the validation step and the download step. The flaw is a classic time-of-check to time-of-use (CWE-367) DNS-rebinding race: app/Rules/PhotoUrlRule.php resolves and approves the submitted hostname, but app/Actions/Import/FromUrl.php later passes the original hostname to app/Image/Files/DownloadedFile.php, which opens a completely separate connection and resolves that hostname again, so an attacker controlling authoritative DNS for the name can return a public address to the check and a private address to the fetch. Exploitation requires a valid Lychee account (PR:L) and precise timing against the second resolution (AC:H), and real-world impact is bounded by which internal services the Lychee host can actually reach and by TLS behavior on the fetched connection, as the vendor advisory notes; no public exploit code has been identified at time of analysis and there is no confirmed active exploitation (CISA KEV).
SGLang up to and including 0.5.19 can be crashed through its session-control API: an authenticated client that issues a generation request with session_params.replace=true and rid=None against a session that already holds request nodes triggers the cleanup loop in Session.create_req, which mutates the request-node dictionary while iterating it and raises an unhandled exception. The impact is confined to availability of the affected request/session handling (CVSS 2.1 / assessed CVSS:3.1 A:L) - there is no code execution, privilege escalation, or data exposure - but the flaw is remotely reachable and publicly available exploit code exists, meaning reachable and authenticated SGLang deployments are at risk of request-path crashes. The upstream fix sits in an unmerged pull request, so no released patched version is confirmed and operators must rely on exposure reduction until the fix lands.