QR Redirector WordPress plugin through 2.0.5 allows an authenticated low-privilege user to store malicious input via its redirect-entry functionality, which then executes in the browser of a higher-privileged victim who views the affected page. This is a moderate-severity stored cross-site scripting issue tracked as CVE-2026-105069 (CVSS 6.5, CWE-79); exploitation requires an authenticated account and victim interaction, and no public exploit identified at time of analysis.
Cross-tenant issue manipulation in Plane (open-source project management tool) before 1.4.0 lets an authenticated user add issues from any workspace into a cycle they control by supplying arbitrary issue UUIDs, and when a victim issue already belongs to another cycle the operation reassigns it, silently removing it from the victim's cycle. The flaw is a user-controlled-key authorization bypass (CWE-639/IDOR) in the cycle-issues endpoint, scored CVSS 6.5 (AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:N) because it requires a valid account (PR:L) and requires the attacker to know or enumerate victim issue UUIDs, which are not trivially guessable and constrain opportunistic abuse. Vendor-released patch: 1.4.0. No public exploit identified at time of analysis; realistic impact is concentrated on multi-tenant or shared Plane deployments where untrusted users share an instance.
Plane's project member roster API (GET /api/v1/workspaces/{slug}/projects/{project_id}/members/) leaks the full membership list of a private project - including each member's email address, name, display name, avatar, and role - to any authenticated user who belongs to at least one project in the same workspace, including a Guest. The flaw is in the DRF permission class ProjectMemberPermission, whose SAFE_METHODS branch validated only workspace-scoped project membership without binding the check to the requested project_id, while the view filtered records solely on the URL-supplied project_id. Only Plane versions prior to 1.4.0 are affected, no public exploit code has been identified at time of analysis, and the impact is read-only disclosure of member PII (personally identifiable information) with no write or availability effect; an attacker must already hold a valid account and know or enumerate the target project's ID.
Improper authorization in Plane (makeplane) prior to 1.4.0 allows an authenticated workspace member or guest who is not a member of a victim's secret project to download that project's private file assets through the workspace-scoped endpoint GET /api/assets/v2/workspaces/{workspace_slug}/download/{asset_id}/, which issues a 302 redirect to a signed download URL instead of enforcing project membership. The flaw is a broken-access-control/IDOR issue (CWE-639) rated CVSS 6.5 (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N), with confidentiality-only impact and no integrity or availability effect; exploitation requires a valid workspace-authenticated session plus knowledge of the target FileAsset UUID, and only project-bound asset types (ISSUE_ATTACHMENT, COMMENT_DESCRIPTION, PAGE_DESCRIPTION, PROJECT_COVER) are exposed, while workspace-level assets with a null project_id are unaffected. There is no public exploit identified at time of analysis and the issue is not listed in CISA KEV; a vendor fix is available in Plane 1.4.0 (PR #9372, commit e63f0c3b3404d669ae05dd9050aab72292f87e5c).
SQL injection in the maestro component of Red Hat Multicluster Engine for Kubernetes (MCE) allows remote attackers to perform read-only blind extraction of database contents through the orderBy parameter on REST API list endpoints. The vendor CVSS vector specifies PR:L, indicating a low-privileged authenticated account may be needed, while the description and independent assessment note that authentication may not be required; defenders should resolve this discrepancy because it changes exposure materially. At time of analysis there is no public exploit identified, and exploitation is constrained by blind, read-only extraction and the requirement for network reachability to the vulnerable MCE API; CVSS is 6.4 medium severity.
Server-side request forgery (SSRF) in Event Gallery for Joomla before 6.6.0 allows an attacker to make the server send the Google Photos OAuth access token to an arbitrary address or probe internal network resources. The vulnerability resides in the back-end upload page's Google Photos picker, which fetches image thumbnails using a user-supplied address without validation and without a form token, enabling exploitation via CSRF against a logged-in administrator or directly by a back-end user with 'Manage' permission. No public exploit code has been identified, and the risk is assessed as moderate and targeted.
HTTP request smuggling in Progress Telerik Fiddler Classic for Windows (versions prior to v6.0.20262.10021) allows a low-privileged local user sharing the same Fiddler proxy instance to desynchronize requests and poison responses delivered to other users. Fiddler forwards conflicting multiple Content-Length headers verbatim to the origin while framing the body using only the first value, and because the server connection is returned to Fiddler's pipe pool after reading only the first response, an unread smuggled response stays buffered on the socket and is served to the next session that reuses it. Exploitation is authenticated (CVSS vector PR:L), local-only (AV:L), and high complexity (AC:H), requiring a non-RFC-9110-compliant keep-alive origin server; no public exploit code or confirmed active exploitation was identified at time of analysis, and the vendor has released a fix.
apko truncates out-of-range UID and GID values that it parses from package-supplied /etc/passwd and /etc/group entries, so a crafted package can make an account that appears unprivileged be written into the resulting image as UID 0 / GID 0 (root). An attacker must first control a package installed into the image being built - a supply-chain foothold rather than arbitrary remote access - and the impact is confined to the produced image's file ownership and its resolved run-as user, not the build host itself. The flaw affects apko versions 0.2.0 up to but not including 1.4.5 on 64-bit platforms; it is rated 6.3 under CVSS 4.0 (independently assessed as AV:N/AC:H/PR:N/UI:N/S:C/C:N/I:L/A:N on CVSS 3.1), and no public exploit code or confirmed active exploitation was identified at time of analysis. Vendor-released patch: 1.4.5.
A double-free flaw in MediaTek's AI Processing Unit (APU) component allows a locally executing attacker - such as a malicious or compromised installed application - to crash an affected device, producing a local denial of service. Per the assessed CVSS vector (AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H), no authentication, elevated privileges, or user interaction are required once local code execution is achieved, and the impact is limited to availability; there is no data exposure or code-execution gain from this defect. Exploitation is confined to devices built on the enumerated MediaTek chipsets (MT8781, MT6899, MT8668, MT8910, MT6993) running unpatched firmware; no public exploit code has been identified at time of analysis and the issue is not listed as confirmed actively exploited (CISA KEV).
Path traversal in the Totolink A3002MU router's File Upload Handler (firmware 1.0.0-B20230403.1455) allows remote unauthenticated attackers to write files outside the intended upload directory by manipulating the 'filename' argument to /boafrm/formUploadFile. Publicly available exploit code exists, but there is no confirmed active exploitation (CISA KEV). The CVSS 4.0 score is 5.3 with low confidentiality, integrity, and availability impact, and the vector requires passive user interaction (UI:P), so exploitation is not fully unattended. Realistic risk depends on whether the upload interface is reachable from untrusted networks.
PHP object injection in FacturaScripts before version 2026.7 allows an authenticated attacker to delete arbitrary files on the server, causing denial of service and potentially enabling a reinstall hijack that resets the application. The flaw is reachable through the multiple-select handling path in WidgetSelect::processFormData(), which passes raw POST data to unserialize() without restricting allowed classes; submitting a crafted serialized XLSXWriter object triggers its __destruct() destructor, which performs the file deletion (config.php and backup data are the notable targets). Exploitation requires a valid authenticated session (CVSS 4.0 vector reports PR:L, so an account is mandatory), and no public exploit code has been identified at time of analysis, so priority is moderate rather than urgent. The vendor has released the fixed 2026.7 version.
Reflected cross-site scripting in FeelCRM OS 1.0.0 allows unauthenticated remote attackers to inject arbitrary script through the keyword parameter in the Department Search endpoint handled by GroupController::index, requiring a victim to visit a crafted link. Publicly available exploit code exists, but there is no confirmed active exploitation (not in CISA KEV); the independent assessment rates this as low severity with vendor CVSS 4.0 base score 2.1 and assessed CVSS 3.1 vector AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:N/A:N. The primary risk is script execution in the victim's browser context, especially if the victim holds an authenticated CRM session.
Code injection in TallCMS up to and including 4.8.0 lets an authenticated admin-panel user who is not a super admin upload and install a plugin ZIP containing arbitrary PHP that is then autoloaded and executed by the application. Exploitation requires valid credentials at a privilege level below super_admin plus the plugin-upload capability to be enabled (the uploadsAllowed() config flag), so it is not reachable anonymously, but publicly available exploit code exists per the referenced GitHub issue and no CISA KEV entry indicates confirmed active exploitation. Note that the vendor/CVSS 4.0 score (5.3, low confidentiality/integrity/availability impact) substantially understates the issue relative to the independent assessment (CVSS 3.1 8.8, C:H/I:H/A:H), which treats the arbitrary-PHP execution as full compromise of the installation; the signals across sources are mixed and partly contradictory.
Path traversal in the Gerapy project management component allows an authenticated attacker to write arbitrary files via a crafted project_name or malicious ZIP archive upload, affecting versions up to 0.9.13. Publicly available exploit code exists, but there is no evidence of active exploitation at time of analysis. The vulnerability has a moderate CVSS score of 5.3 and requires an authenticated session (CVSS PR:L).
Authenticated SQL injection in feelec-yishu feelcrm-os 1.0.0 lets a low-privileged logged-in user run attacker-controlled SQL through the groups[] array parameter handled by the getMemberByGroups AJAX endpoint in App/Feelcrm/Crm/Controller/AjaxRequestController.class.php. Publicly available exploit code exists (the reporter published details in the project's GitHub issue tracker at https://github.com/feelec-yishu/feelcrm-os/issues/1), but there is no indication of confirmed active exploitation and the flaw is not in CISA KEV. The vendor's CVSS 4.0 score is 2.1 while an independent assessment rates it a genuinely low-priority, remotely reachable SQLi constrained by mandatory valid credentials (CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:N), no user interaction, and the product's rare self-hosted deployment footprint.
SQL injection in FeelCRM OS 1.0.0 allows an authenticated, low-privileged remote attacker to execute arbitrary SQL by manipulating the group_id argument handled by the index function of App/Feelcrm/Index/Controller/MemberController.class.php, exposing or altering member-related data. Publicly available exploit code exists (the CVSS 4.0 vector carries E:P, and the original report is a public GitHub issue), but there is no evidence of confirmed active exploitation (CISA KEV). Because exploitation requires a valid low-privileged account (CVSS PR:L) and yields only low confidentiality, integrity, and availability impact, this is a low-to-moderate priority issue rather than an unauthenticated critical flaw.
Session fixation in the kishor-23 food-waste-management-system PHP application lets unauthenticated remote attackers take over a logged-in user's session by fixing the PHPSESSID value before the victim authenticates through login.php, because the application never regenerates the session identifier after a successful login. Exploitation requires induced victim interaction (UI:P/UI:R) - the victim must log in while the attacker-controlled or attacker-predicted session ID is active - and impact is confined to that single user's session rather than the underlying host. Publicly available exploit code exists (GitHub issue #16), but this is an unversioned, niche GitHub project with no vendor response and no released patch, so despite the moderate 5.3 CVSS score the practical deployment risk is genuinely low.
Unrestricted file upload in feelec-yishu FeelCRM-OS 1.0.0 allows an authenticated, low-privileged user to write arbitrary files to the server through the UploadTicketFile endpoint at App/Feelcrm/Crm/Controller/UploadController.class.php by manipulating the cmd argument. The issue is reachable remotely over the network, requires no user interaction, and publicly available exploit code exists, but exploitation is gated on holding a valid low-privileged account (CVSS PR:L) and is confined to version 1.0.0 of this narrowly deployed CRM product, so overall risk is low. The vendor/project was notified through an issue report and has not responded, and there is no indication of confirmed active exploitation (no CISA KEV listing).
Authenticated SQL injection in itsourcecode Online Admission System 1.0 lets a user who can reach the administrative interface inject arbitrary SQL through the 'sy' (school year) parameter of /admin/schoolyear.php, allowing read and modification of backend database contents. The attack is network-reachable but requires a valid session on the admin panel (the assessed vector is CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L), so unauthenticated internet attackers cannot reach the vulnerable endpoint without first obtaining or bypassing admin credentials. Publicly available exploit code exists (a proof-of-concept has been published via the vendor-adjacent GitHub issue linked from VulDB), but there is no confirmed active exploitation (CISA KEV) and the CWE-89 flaw is assessed as low priority. EPSS data was not provided in this intelligence set.
SQL injection in itsourcecode Online Admission System 1.0 allows an attacker who already holds a valid low-privilege session to manipulate the ID parameter of /admin/key.php and inject arbitrary SQL into the backend database. Exploitation is remote but gated behind authentication and access to the administrative endpoint, so it is a low-priority, low-real-risk issue per our assessment; the practical exposure is confined to insiders or attackers who have already obtained credentials. Publicly available exploit code exists, no public exploit has been confirmed as actively exploited, and no CISA KEV entry is associated with this CVE.
SQL injection in itsourcecode Online Admission System 1.0 lets an authenticated, low-privileged user manipulate the schedid argument on /new.php so that attacker-controlled input is interpreted as part of a backend SQL statement, giving limited confidentiality, integrity and availability impact with no user interaction required. The issue is remotely reachable and publicly available exploit code exists (a public GitHub issue demonstrates the vector), but CISA KEV does not list it and no vendor-released patch was identified in the available data, so this remains a moderate-severity, low-priority issue for most deployments. Exposure is bounded by the authentication prerequisite: an attacker needs a valid low-privileged session in the application, and from the reported vector the injection does not inherently yield full database or host control.
SQL injection in itsourcecode Online Admission System 1.0 allows remote attackers to read or alter database contents via the fname argument processed by register1.php. The flaw is tracked as CVE-2026-105181 (CWE-89) with a CVSS 4.0 base score of 5.3 and publicly available exploit code; there is no CISA KEV confirmation of active exploitation. The CVSS vector indicates PR:L, but because register1.php is a registration endpoint, the required privilege level is questionable and may in practice be unauthenticated - defenders should verify.
Stored cross-site scripting in Bacularis 5.4.0 through 6.5.1 allows an attacker to write script content into the "Organization name" field, which is then rendered unsafely and executes in the browser of any user who views the affected page. The independent assessment rates this moderate at CVSS 6.1, and it requires an authenticated account holding organization create/edit rights plus victim interaction - a separate user, typically an administrator, must load the view that renders the stored name - with impact confined to the browser session and no direct server compromise. The supplied CVE record vector lists PR:N, but the assessed vector and exploitation conditions indicate PR:L is correct; exploitation status is no public exploit identified at time of analysis, upstream detection is minimal, and EPSS is low at 0.15% (4th percentile), though the 6.1 score together with a scope-changing (S:C) XSS in an administrative interface keeps this relevant for defenders who have not yet reviewed stored organization data.
Stored cross-site scripting in FeelCRM-OS 1.0.0 lets an authenticated low-privileged user inject arbitrary script through the customer_form[remark] field of the Create Customer endpoint, which is persisted after htmlspecialchars_decode processing and executes when another user views the affected customer record. Publicly available exploit code exists (referenced through the vendor's GitHub issue tracker and VulDB), but there is no confirmed active exploitation via CISA KEV and no vendor-released patch has been identified. The issue is a low-to-moderate priority item because exploitation requires valid credentials, a second user to render the stored record, and the product is niche with no known default internet-facing deployment.
Server-side request forgery in Penpot before 2.18.0 allows an authenticated user who controls a media-import URL, or an administrator who controls a webhook URL, to reach internal and cloud-metadata endpoints by supplying IPv6 transition addresses (NAT64 64:ff9b::/96, 6to4 2002::/16, or Teredo 2001:0000::/32) that embed a blocked IPv4 destination such as 169.254.169.254, 127.0.0.1, link-local, or RFC1918 addresses. The bypass works because the backend blocklist app.util.ssrf/blocked-address? applies its extra CIDR checks only to IPv4 values and relies on Java InetAddress predicates that do not classify transition-mechanism addresses; exploitation further requires the request to traverse a NAT64 gateway or the attacker to control a DNS AAAA record the Penpot server resolves, with cloud environments that already run DNS64/NAT64 directly exploitable, while non-NAT64 deployments depend on the harder attacker-controlled AAAA path. Media import can disclose response bodies, whereas webhook delivery exposes only response status as a network-probing side channel, and there is no public exploit identified at time of analysis. Penpot 2.18.0 is the vendor-released fix (commit 326d83e780ae120b45ccd6f3d7bd24922b54461c).
Remote code execution in Django Haystack's Elasticsearch backend affects versions up to and including 3.3.0, where an attacker-controlled string reaching the result-deserialization path is passed to `eval()` inside `_to_python()` and executed in the Django application process. Exploitation is conditional rather than blanket: it requires the Elasticsearch backend, at least one `SearchField` declared with an `index_fieldname` alias that differs from its logical field name, and the ability to get a malicious payload indexed into that aliased field and returned by a subsequent search - a second-order/stored injection rather than a direct request-driven one. Publicly available exploit code exists (a GHSA advisory with source-level walkthrough was published alongside the fix), and the input CVSS vector marks PR:H, so under that vector exploitation is treated as authenticated; there is no CISA KEV entry indicating confirmed active exploitation. The vulnerability is patched in 3.4.0, and the reporting signals conflict sharply and must be weighed carefully: VulDB's CVSS 4.0 base of 2.0 contrasts with the upstream advisory's CVSS 3.1 base of 8.5, reflecting the gap between the mild vector assumptions and the full RCE impact when the alias configuration is present.
BuildKit (moby/buildkit) contains a policy-validation bypass in which a specially crafted Git source build step makes a repository appear to originate from a different remote URL than the one actually used during git clone, so repository-identity or URL-based policy rules can be evaded. The issue affects BuildKit builds prior to v0.33.1 and requires an attacker to hold build-submission privileges (CVSS vector PR:L); policies that validate commit SHA, commit data, or cryptographic signatures continue to apply correctly, which limits the practical impact (vendor CVSS 4.0 base 6.0, independent assessment 3.1 base with AC:H, low confidentiality and high integrity impact, no availability impact). No public exploit code and no confirmed active exploitation (CISA KEV) were identified at time of analysis; the vendor reports a patch is available.
Cross-site scripting in code-projects HRMS 1.0 allows an authenticated user with admin-area access to inject script via the eventSubject parameter in the Live Event History page, and the payload executes in a victim's browser when that page is viewed. Publicly available exploit code exists, but exploitation requires admin-area access and a separate victim to load the affected page; the assessed impact is integrity-only with no confidentiality or availability loss. This is a low-to-moderate real-world risk that scores higher on paper than in practice.
Stored cross-site scripting (XSS) in code-projects Human Resource Management 1.0 lets an authenticated, low-privileged user inject arbitrary JavaScript through the eventSubject argument handled by /humanresourcemanagementsystem/src/store/EventStore.php, which then executes in the browser of any user who views the affected event. Exploitation requires an account able to reach the Event Creation feature plus a victim who renders the malicious event (UI:P), and publicly available exploit code exists; there is no confirmed active exploitation (not in CISA KEV). Independent assessment rates real-world risk as low because of the authentication prerequisite, the dependence on victim interaction, the scope change toward other users in the same origin, and the very small deployment base of this product.
Login cross-site request forgery in the Chaterm desktop client before 0.12.1 allows unauthicated remote attackers to inject a forged login state by triggering a crafted chaterm:// OAuth deep-link callback that the client neither signs nor validates against the original state parameter. A victim only needs to visit an attacker-controlled page while the chaterm:// handler is registered; the client then completes login as the attacker's account, and if default data sync is enabled, saved hosts, passwords, and private keys are uploaded into that attacker-controlled account. Exploitation is conditional rather than trivial - it requires victim interaction and a reachable callback/login state (AC:H, AT:P), and the confidentiality impact scales with how many secrets the victim has stored; there is no confirmed active exploitation and no public exploit code identified at time of analysis, and the vendor has released the fix in 0.12.1, which adds RSA-SHA256 signature verification of deep-link callbacks plus time-bound pending-state tracking.
BuildKit, the container build engine used by Docker and exposed through `docker build`/buildx, can be tricked into performing file operations on special file inodes - device nodes, FIFOs, or sockets - at paths where it expects ordinary regular files, a confused-deputy condition (CWE-441). A user who builds an attacker-controlled Dockerfile or build context can therefore cause build operations to hang/block, and on rootful workers can obtain unintended access to host devices; rootless configurations are effectively limited to availability and integrity effects. The attack is local and requires the victim to actively run a build on the malicious input, so exposure concentrates in multi-tenant or untrusted-source CI/CD pipelines that use rootful BuildKit workers; no public exploit has been identified at time of analysis and the flaw is not in CISA KEV. This is best treated as a medium-severity, defense-in-depth hardening issue rather than an urgent remote-exploitation priority.
Unauthenticated retrieval of customer-uploaded files is possible in the File Uploads Addon for WooCommerce WordPress plugin (versions 1.7.2 through 1.7.6), which stores customer submissions in a publicly web-accessible uploads directory while the access restriction it generates fails to actually block direct requests; an attacker who knows or can guess a file's name can fetch it straight from the web root, bypassing the plugin's authenticated download mechanism. The issue is a broken access control / insecure direct object reference problem (CWE-284) rather than a true authentication bypass, and only confidentiality of already-uploaded files is affected - no integrity or availability impact. Rated CVSS 5.9 (AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N), the attack is unauthenticated and needs no user interaction, but the High attack-complexity component reflects that the attacker must correctly guess or otherwise learn the target filename; there is no public exploit identified at time of analysis and no evidence of confirmed active exploitation (CISA KEV).
Unauthenticated disclosure of other customers' uploaded files is possible through the File Uploads Addon for WooCommerce WordPress plugin in versions 1.7.2 through 1.7.5, because the plugin serves stored uploads without verifying that the requester is the customer who originally uploaded the file (CWE-639). An attacker who knows or guesses the exact file name of a previously uploaded file can retrieve it over the network with no account and no user interaction, as reflected by CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N (5.9 Medium) - the High attack complexity stems entirely from the file-name discovery prerequisite, so real-world difficulty hinges on whether the plugin generates guessable or sequential names versus long random tokens. There is no public exploit identified at time of analysis, and the vendor has released a fix in 1.7.6.
Code injection in osCommerce2 (versions through 2.3.4.1) allows an attacker who already holds high backend privileges - a store administrator able to edit the payment configuration - to manipulate the MODULE_PAYMENT_INSTALLED parameter and have attacker-controlled content pulled into a PHP include() in includes/classes/payment.php. Because the CVSS 4.0 vector carries PR:H and UI:P, the vulnerable payment flow is not reachable by an anonymous shopper and requires at least passive user interaction, which keeps this a medium-severity, post-authentication issue (base score 4.8) rather than a pre-auth remote takeover. Publicly available exploit code exists, and there is no confirmed upstream fix because the project has not responded to the reporter's early notification; no EPSS or CISA KEV signal is available for this CVE.
BuildKit can cache a fetched snapshot under a claimed content digest without adequately verifying that the blob bytes match that digest, letting an attacker who controls or compromises a registry or OCI-layout blob source poison the cache and compromise the integrity of later victim builds that reuse that digest. The attacker does not need credentials per the CVSS PR:N vector, but they must be able to serve BuildKit blob requests, and a victim must run a build against the attacker-influenced source for exploitation to matter. No public exploit was identified at time of analysis; vendor patch v0.33.1 is available, and an independent assessment rates this as a moderate, conditions-dependent integrity flaw rather than a drop-everything emergency.
Penpot versions prior to 2.18.0 leak the contents of other users' in-progress uploads because the backend assemble-chunks routine fetches an upload session by ID alone instead of scoping the lookup to the calling account. An authenticated attacker (PR:L) who comes to know a victim's live, completed-but-not-yet-assembled upload-session UUID can assemble the victim's stored chunks into the attacker's own file, team font, or project import - reading the uploaded bytes (C:H) and wiping the victim's pending session (A:L). Successful abuse is gated by the UUID's randomness and the brief window before the session is consumed, which the assessed vector captures as AC:H; no public exploit code or CISA KEV entry is associated with this issue at time of analysis.
Logout in Penpot versions prior to 2.18.0 clears the browser's auth-token cookie but leaves the matching server-side session record in the http_session_v2 table intact, so a previously captured session token continues to authenticate requests with the victim's authority until it naturally expires. This is a medium-severity session-management flaw (CVSS 5.9, CWE-613) rated with a network vector, no privileges required (PR:N), and user interaction (UI:R) - although in practice an attacker must first obtain a valid token through a separate vector such as a shared machine, interception, or XSS, and only that specific captured session is affected. The vendor addressed the issue in the 2.18.0 release and via fixing commit 7c85837, and no public exploit code has been identified at time of analysis.
Server-side request forgery in Docling 2.91.0 through 2.131.x lets an attacker-supplied document reach internal services whenever remote document fetching is explicitly enabled, because validate_url_safety in docling/backend/utils/image_resource_loader.py performs only a single IPv4 lookup before the HTTP client re-resolves and parses the original URL. The same trust gap applies to HTMLBackendOptions(render_page=True), which issues HTTP and HTTPS browser requests without validating their resolved destination, so exploitation also covers DNS rebinding, mixed public/internal address records, and backslash-authority parser disagreement. Exploitation is gated by non-default configuration (fetch_images=True with enable_remote_fetch=True, --html-image-fetch remote, or render_page=True), requires the attacker to supply a processed document and win a rebinding timing race, and exposure is limited to responses successfully decoded as images or passively captured in a page screenshot; CVSS 3.1 is 4.0 (AV:N/AC:H/PR:N/UI:N/S:C/C:L/I:N/A:N), no public exploit identified at time of analysis, and the issue is fixed in 2.132.0.
Moby BuildKit (all versions in the affected range per the unbounded CPE cpe:2.3:a:moby:buildkit:*:*:*:*:*:*:*:*) can be made to run its cleanup routine against a modified CA bundle when proxy networking with CA injection is enabled, a time-of-check-to-time-of-use (CWE-367) race that causes cleanup to block, operate outside the build rootfs, or fail without failing the build. The attacker must be able to submit or execute a build on the builder (local, low-privilege access; PR:L) and must first change the build's injected CA bundle before cleanup runs; the non-default proxy-with-CA-injection mode sharply narrows exposed deployments, and the dominant consequence is availability and integrity of the cleanup operation (VA:H, VI:L) rather than confidentiality loss. No public exploit code was identified at time of analysis and the issue is not in CISA KEV; vendor-released patch v0.33.1 is available.
Stored cross-site scripting in Zabbix lets a crafted email media type configuration run arbitrary JavaScript in the browser of the Super Admin who approves it: the media type's OAuth form passes the attacker-controlled 'Authorization endpoint' value directly to window.open() without validating the URL scheme, so a javascript: URL executes in the frontend's origin. Exploitation is authenticated and interaction-dependent - the payload must first be delivered as an import file by a user with high frontend privileges (PR:H), and a Super Admin must then open that media type's OAuth form and actively click to grant consent (UI:A / UI:R) - which is consistent with the published CVSS 4.0 base score of 5.6 and our assessed CVSS:3.1/AV:N/AC:H/PR:H/UI:R/S:U/C:H/I:L/A:N. No public exploit code and no confirmed active exploitation were identified at time of analysis, so although the flaw is genuine and script execution occurs in a highly privileged session, it is a constrained medium-severity issue rather than a mass-exploitation priority.
Keycloak's User Session Note protocol mapper fails to screen session note keys for sensitive internal credentials, allowing an authenticated delegated client administrator to copy a victim's upstream federated access/bearer token into tokens issued to the application they manage (CVSS 3.1 5.7, AV:N/AC:L/PR:L/UI:R, C:H/I:N/A:N). Exploitation requires the attacker to hold scoped client-administration rights sufficient to configure protocol mappers, a realm federated with external identity providers that persist upstream tokens as user session notes, and the victim to actually authenticate to the attacker-controlled client, so impact is decisive only in federated-token deployments and default configurations without federated token storage are not exposed. Red Hat has published an advisory and Bugzilla record for the issue, but no public exploit code and no confirmed active exploitation were identified in the available intelligence at time of analysis.
A malicious external BuildKit frontend can issue crafted requests against the daemon's internal API that trigger a data race, causing the BuildKit daemon to panic and crash. The issue affects BuildKit deployments where the operator has opted into running an untrusted or attacker-controlled frontend (for example through a `# syntax=` reference to a third-party image); default builds that use trusted first-party frontends are not exposed. It is a medium-severity denial-of-service condition: exploitation requires low-level privileges (PR:L) through the frontend channel, local access (AV:L), and successful timing of a race (AT:P), and there is no public exploit identified at time of analysis. A vendor fix is available in BuildKit v0.33.1.
Missing authentication in the Admin Signup component (admin/signup.php) of the kishor-23 Food Waste Management System allows remote, unauthenticated attackers to manipulate the 'sign' argument and bypass the intended signup/authentication check, resulting in limited confidentiality, integrity, and availability impact on the affected endpoint. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N, E:P) confirms the attack is network-reachable, requires no credentials or user interaction, and that publicly available exploit code exists; there is no CISA KEV entry, so active exploitation in the wild is not confirmed. Real-world risk is modest: the project is an obscure open-source PHP application, generally not deployed at production scale, and the project has not responded to the issue report, meaning no patched release is available.
Authenticated path traversal in the Mitel MiVoice Office 400 web management portal (TCP 443) lets a high-privilege maintenance or administrator user read files outside the intended system logs directory by supplying crafted path input to the Maintenance → System Logs download function. Only users who already hold valid high-privilege portal credentials can trigger it (CVSS vector PR:H), and the impact is limited to confidentiality - files can be read or downloaded, but there is no write, code-execution, or denial-of-service capability. No public exploit code and no confirmed active exploitation were identified at time of analysis, and the assessed risk is moderate rather than critical because exploitation is gated behind authenticated administrative access to an interface that should normally be restricted to trusted networks.
Unauthenticated SQL injection in the signup.php user-registration endpoint of the kishor-23 food-waste-management-system allows remote attackers to manipulate the email, name, and gender parameters to inject arbitrary SQL into the application's database queries. The CVSS vector confirms PR:N/UI:N, meaning no credentials or user interaction are required, and publicly available exploit code exists; however, there is no evidence of confirmed active exploitation (not in CISA KEV), and realistic impact is bounded to Low confidentiality, integrity, and availability for a single application database. Because the project follows a rolling-release model, no affected or patched version numbers are published, and the maintainer has not responded to the upstream GitHub issue reporting the flaw.
SQL injection in itsourcecode Online Admission System 1.0 allows remote attackers to inject arbitrary SQL through the 'ID' parameter of /admin/creteria.php, potentially reading or modifying the underlying database. The vulnerable script is located under the /admin/ path, so although the published CVSS 4.0 vector indicates PR:N (no authentication), real-world exploitation most likely requires either an admin area deployed without access control or an existing administrative session. Publicly available exploit code exists and the flaw was disclosed publicly, but there is no CISA KEV entry and no observed evidence of active exploitation; assessed severity is medium (CVSS 4.0 base 5.5).
SQL injection in itsourcecode Online Admission System 1.0 lets remote attackers inject arbitrary SQL through the schedid request parameter handled by /admin/confirm.php, potentially reading or altering admission-related database records. The vendor's CVSS 4.0 vector (5.5, E:P) claims no authentication is required (PR:N), though the /admin/ path suggests an administrative session may in practice be needed; our independent assessment scores it CVSS 3.1 5.6-adjacent at PR:L with limited confidentiality and integrity impact. Publicly available exploit code exists (a GitHub issue demonstrating the flaw), but there is no indication of confirmed active exploitation and no CISA KEV listing; exposure is limited to internet-reachable deployments of this relatively obscure PHP application.
SQL injection in SourceCodester Drug Recommendation System 1.0 allows unauthenticated remote attackers to execute arbitrary SQL commands via the cmdschool parameter in /Auth/add_student.php. A public exploit has been released (publicly available exploit code exists; no evidence of active exploitation in CISA KEV) and the CVSS 4.0 base score is 6.9 (medium). The vulnerability is limited to low confidentiality, integrity, and availability impact, and real-world risk is low because the application is a niche academic/demo PHP project rarely exposed to the internet.
Read-only administrative users on Red Hat Quay can retrieve repository write (delegate) tokens through the build trigger API, then use those tokens to push arbitrary container images into private repositories despite their restricted role. Exploitation requires an account already holding global read-only administrative privileges (CVSS vector PR:H) plus repositories that have build triggers configured with embedded delegate tokens, so realistic risk is confined to malicious insiders or attackers who have already compromised a read-only admin credential. The flaw is rated moderate at CVSS 5.5 (CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:L/I:H/A:N), and no public exploit identified at time of analysis.
Privilege-model bypass in Rapid7 Velociraptor's Windows.Sysinternals.SysmonLogForward artifact allows an authenticated operator who holds only the COLLECT_CLIENT permission (typically the built-in 'Investigator' role) to launch an arbitrary, attacker-chosen binary on monitored Windows endpoints instead of the installed Sysmon binary. Because the artifact exposes a user-controllable binary-path parameter and fails to enforce the additional EXECVE permission that gates comparable dangerous artifacts, successful collection results in code execution on the endpoint under the elevated privileges the artifact normally runs with, yielding high confidentiality, integrity, and availability impact (assessed CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H). Exploitation is not anonymous: the attacker must already be a trusted Velociraptor operator with artifact-collection rights, and the chosen binary must be reachable or placeable on the target endpoint, which narrows realistic exposure to insiders or compromised operator accounts. No public exploit code or confirmed active exploitation was identified at time of analysis; the issue is a moderate, privilege-escalation class flaw rather than a mass-exploitation threat.