| CVE-2025-67649 | A SQL injection vulnerability has been identified in PHP Jabbers - Car Rental Script . Improper neutralization of input provided by user into parameters responsible for sorting functions allows an unauthenticated attacker to perform SQL Injection attacks.
This issue was fixed in version 4.1. |
| CVE-2026-14540 | A Server-Side Request Forgery (SSRF) vulnerability exists in the generic HTTP source and tool components of Google mcp-toolbox versions 0.3.0 through 1.4.0. While the toolbox implements baseline input sanitization for user-controlled parameters, the underlying HTTP client (internal/sources/http/http.go) fails to safely regulate request redirection boundaries. Specifically, the client is initialized without a restrictive CheckRedirect policy hook and lacks target IP validation. An attacker or a malicious data-driven prompt can supply a crafted path parameter that triggers an open redirect or a direct destination swap on the target backend, coercing the mcp-toolbox into blindly following the redirection and making unauthorized requests to internal or arbitrary external endpoints. |
| CVE-2026-14541 | An authentication bypass and audience confusion vulnerability exists in the Google OAuth provider component of Google mcp-toolbox version 1.4.0. When a Google authService is initialized with mcpEnabled: true but lacks an explicitly defined audience or clientId, the ValidateMCPAuth pipeline for opaque tokens skips audience validation entirely. As a result, the toolbox will accept any valid Google OAuth access token—even those minted for unrelated ecosystem applications—granting unauthorized clients access to protected tools and data backends. |
| CVE-2026-14922 | WP Photo Album Plus is vulnerable to stored Cross-Site Scripting in all versions up to, and including, 9.2.03.001 through a decode-after-sanitize (double-encoding) flaw in the photo-comment pipeline. On write, `wppa_do_comment()` sanitizes the comment with `wppa_filter_html()` (wp_kses) followed by `wp_strip_all_tags()` (`wppa-functions.php:2623-2624`). Because `wp_strip_all_tags()` only removes *real* tags, an attacker who submits a **double HTML-entity-encoded** payload (e.g. `&lt;img src=... onload=...&gt;`) passes the write filters as harmless entity text and is stored one decode-level down (`<img ... onload=...>`). |
| CVE-2026-15722 | A stack buffer overflow flaw was found in 389 Directory Server (389-ds-base). The get_ruvelement_from_berval() function in repl5_ruv.c copies digit characters from a network-supplied RUV berval into a fixed 16-byte stack buffer without bounds checking. A remote unauthenticated attacker can crash the LDAP server by sending a crafted StartNSDS50ReplicationRequest extended operation containing a replica ID field with more than 16 digit characters. The overflow occurs during payload decoding, before any authorization check. Stack protectors limit impact to denial of service. |
| CVE-2026-16105 | A flaw was found in the RoleContainerResource component of Keycloak. The issue occurs because certain name-based endpoints in the admin REST API do not properly enforce authorization checks when managing composite roles. This allows a delegated administrator with manage-realm permissions to remove essential child roles from built-in admin roles, potentially disrupting administrative functions within a realm. |
| CVE-2026-16503 | Deployment of the VPS.org one-click Supabase template deploys a PostgreSQL instance that is published on all interfaces (0.0.0.0:5432) with a default database password set to "postgres". Because Docker installs its own iptables rules, this exposure bypasses a standard host UFW configuration. |
| CVE-2026-16504 | Deployment of the VPS.org one-click Zulip template deploys a hardcoded application signing key, a default database password ("zulip"), and DISABLE_HTTPS=True. |
| CVE-2026-17561 | Improper Control of Generation of Code ('Code Injection') vulnerability in Innotim Software, Telecommunications and Consulting Trade Ltd. Co. Logsign SIEM allows Code Injection.
This issue affects Logsign SIEM: before 6.4.108. |
| CVE-2026-17566 | pgAdmin 4's Import/Export Data tool builds a psql \copy (...) command line by interpolating a user-supplied SQL query into a Jinja template and passing the rendered line to psql via --command. To stop an attacker from breaking out of the (...) wrapper, create_import_export_job() (route POST /import_export/job/<sid>, gated only by the ordinary, commonly-granted tools_import_export_data permission) validated the query with a hand-written parenthesis-balance checker, _is_query_parens_balanced(). That checker always treated a backslash before a single quote (\') as escaping the quote, i.e. as if standard_conforming_strings were off. PostgreSQL has defaulted standard_conforming_strings to on since 9.1 (2010), the default on every PostgreSQL version pgAdmin 4 currently supports (13-18); under that default psql's own \copy tokenizer treats \ as an ordinary character, so a single quote immediately after it closes the string literal. A query such as SELECT 'a\') TO PROGRAM 'echo pwned' x' was therefore accepted as "balanced" by pgAdmin's checker (which believed the ) was still inside the string), while psql, run through the actual rendered command line, closes the string at that point and treats the following ) as the end of the wrapping \copy (...) subquery, exposing an attacker-chosen TO PROGRAM '<command>' clause that psql executes via popen() -- independent of a subsequent syntax error later on the same line. This is the same class of bug as CVE-2025-12762/CVE-2025-13780 (RCE via psql meta-command/COPY injection during PLAIN-format dump restore), reached through an independently written defense in a different module (Import/Export Data rather than Restore) that had its own, different logic bug (inverted backslash-escape semantics rather than a BOM-defeated regex anchor).
The fix rejects any backslash inside a single-quoted string in the query outright, rather than picking one of the two possible psql interpretations. This is intentionally conservative: because the correct interpretation of \ depends on the target server's standard_conforming_strings setting, which the checker cannot reliably know at validation time, refusing the query is safer than guessing.
This issue affects pgAdmin 4: from the introduction of _is_query_parens_balanced() before 9.18. |
| CVE-2026-18203 | A flaw was found in the group policy evaluation logic of Keycloak, an identity and access management solution. When a group policy is set to extend permissions to child groups, the system incorrectly uses a simple text-based prefix check to verify group membership. This allows a user who belongs to a different group with a similar starting name to bypass security checks and gain unauthorized access to administrative functions or protected resources. |
| CVE-2026-18206 | A flaw was found in the keycloak-services component of Keycloak, which provides identity and access management services. The issue occurs when a realm administrator uses a wildcard domain (like *.example.com) to restrict which hosts can register or update clients. Due to improper validation, the system accepts any hostname that ends with the specified domain suffix, even if it is not a legitimate subdomain. An attacker who can control the reverse DNS of their connection can bypass these host-based restrictions, potentially allowing unauthorized client modifications. |
| CVE-2026-18208 | A flaw was found in the OIDC token introspection endpoint of the keycloak-services component. Keycloak is an open-source identity and access management solution used to secure modern applications and services. The issue occurs when a confidential client, configured to receive signed JWT introspection responses, attempts to introspect a token issued for a different audience. Although the endpoint correctly identifies the token as inactive for that client, it still returns the full set of token claims within a signed JWT field. This allows an unauthorized client to bypass audience-based restrictions and access sensitive information contained in the token. |
| CVE-2026-18209 | A flaw was found in the keycloak-services component of Keycloak, which handles OpenID Connect (OIDC) authentication flows. The issue occurs because the security check designed to prevent HTTP parameter pollution only inspects the query portion of a redirect URL and ignores the fragment portion. When a client is configured with a wildcard redirect URI, an attacker can use this to inject duplicate security parameters into the login response. If a client application is not configured correctly, it might trust the attacker's injected data instead of the real security information from Keycloak, leading to session fixation or account confusion. |
| CVE-2026-18211 | A flaw was found in the secure-client-uris client policy executor within Keycloak core services. This component is responsible for enforcing security requirements on client configurations, such as requiring encrypted connections for redirect URIs. Due to an improper check that only looks at the start of a web address rather than properly verifying the host, an attacker can bypass these security restrictions by using a specially crafted domain name. This could allow an attacker to intercept sensitive authentication codes over unencrypted connections. |
| CVE-2026-18214 | Keycloak allows users to log in using Google accounts and can be configured to only allow users from specific Google Workspace domains. A flaw was found where the token exchange feature, which allows swapping a Google token for a Keycloak token, does not check these domain restrictions. This means an attacker with a valid Google account from a different domain could bypass the security check and gain access to the Keycloak realm. |
| CVE-2026-18215 | Keycloak provides a way to let users log in using Microsoft accounts while restricting access to a specific organization (tenant). A flaw was discovered where this restriction is ignored when using the token exchange feature. This means an attacker with a valid Microsoft token from a completely different organization could gain access to the Keycloak realm, potentially accessing sensitive data or performing unauthorized actions. |
| CVE-2026-18217 | A flaw was found in the SAML protocol implementation of Keycloak, an open-source identity and access management solution. The issue occurs when Keycloak handles SAML authentication requests using the HTTP-Redirect binding. If a client is configured with a wildcard redirect URL, an attacker can craft a request that includes malicious parameters. When a user authenticates, Keycloak appends its legitimate response to the attacker's parameters. This can cause some service providers to process the attacker's data instead of the real login information, potentially leading to a user being logged into the wrong account. |
| CVE-2026-18218 | A flaw was found in the TokenManager component of the Keycloak identity management service. When an administrator attempts to revoke tokens for a specific application (client) using a "not-before" policy, the revocation may be silently ignored if the overall security realm already has an older, non-zero revocation policy in place. This issue can allow previously issued tokens to remain valid for refreshing sessions and accessing user information even after an administrator has attempted to invalidate them.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ |
| CVE-2026-18321 | Buffer overflow in NTPsec's Zyfer refclock allows local attacker to crash ntpd |
| CVE-2026-18446 | fast-uri before 4.1.2, 3.1.5, and 2.4.4 requires a literal double forward slash to recognize a URI authority, so a reference that uses a backslash based introducer in place of it (backslash backslash, forward slash backslash, or backslash forward slash) is parsed with no authority and folds into the path. Node's native WHATWG URL parser instead treats a backslash as interchangeable with a forward slash for special schemes, so the two parsers extract different hosts from the same input. Applications that use fast-uri to enforce host based policy such as allowlists, SSRF filtering, or redirect validation before passing the same URL into Node's URL or fetch consumers can be steered to an unintended host. Upgrade to fast-uri 4.1.2, 3.1.5, or 2.4.4. |
| CVE-2026-18452 | DMS+ (Non-Mobile) developed by Rich Source has a Use of Hard-coded Credentials vulnerability. Unauthenticated remote attackers can exploit a fixed API key to gain control over all installed DMS+ devices. |
| CVE-2026-21662 | Unrestricted upload of file with dangerous type vulnerability in Johnson Controls FM Systems Employee allows Using Malicious Files.
This issue affects FM Systems Employee: before 2025.3.1. |
| CVE-2026-28144 | Insertion of Sensitive Information Into Sent Data vulnerability in Flipper Code WP Maps allows Retrieve Embedded Sensitive Data.
This issue affects WP Maps: from n/a through 4.9.6. |
| CVE-2026-28145 | Insufficient Verification of Data Authenticity vulnerability in StylemixThemes MasterStudy LMS allows Manipulating User State.
This issue affects MasterStudy LMS: from n/a through 3.7.39. |
| CVE-2026-34641 | Premiere Pro is affected by an out-of-bounds write vulnerability that could result in arbitrary code execution in the context of the current user. Exploitation of this issue requires user interaction in that a victim must open a malicious file. |
| CVE-2026-38708 | TR1200 v2.4.15, TR3000 v2.4.21, WR300 v2.4.25, WR1200 v2.4.23, WR1300 v2.4.22, WR1500 v2.3.10, WR3000 v2.4.19, WR3600 v2.3.16, and WR6500 v2.3.15 were discovered to contain a command injection vulnerability in the system.setclock interface. This vulnerability allows attackers to execute arbitrary commands as root via a crafted input. |
| CVE-2026-38710 | TR1200 v2.4.15 and TR3000 v2.4.21 were discovered to contain a command injection vulnerability in the system.setclock interface. This vulnerability allows attackers to execute arbitrary commands as root via a crafted input. |
| CVE-2026-38711 | TR1200 v2.4.15, TR3000 v2.4.21, WR300 v2.4.25, WR1200 v2.4.23, WR1300 v2.4.22, WR1500 v2.3.10, WR3000 v2.4.19, WR3600 v2.3.16, and WR6500 v2.3.15 were discovered to contain a command injection vulnerability in the system.upgrade_check interface. This vulnerability allows attackers to execute arbitrary commands as root via a crafted input. |
| CVE-2026-38713 | TR1200 v2.4.15, TR3000 v2.4.21, WR300 v2.4.25, WR1200 v2.4.23, WR1300 v2.4.22, WR1500 v2.3.10, WR3000 v2.4.19, WR3600 v2.3.16, and WR6500 v2.3.15 were discovered to contain a command injection vulnerability in the ipsec_conn interface. This vulnerability allows attackers to execute arbitrary commands as root via a crafted input. |
| CVE-2026-43829 | Full details and mitigation steps are currently restricted and will be published at a later date. |
| CVE-2026-43830 | Full details and mitigation steps are currently restricted and will be published at a later date. |
| CVE-2026-43831 | Full details and mitigation steps are currently restricted and will be published at a later date. |
| CVE-2026-43832 | Full details and mitigation steps are currently restricted and will be published at a later date. |
| CVE-2026-43833 | Full details and mitigation steps are currently restricted and will be published at a later date. |
| CVE-2026-44615 | Path traversal vulnerability in Apache Zeppelin. When FileSystemNotebookRepo is configured, an authenticated attacker with permission to rename a note, or access to folder operations, could supply traversal segments in note or folder paths. Zeppelin composed these values into filesystem paths using the server's filesystem or Hadoop identity without ensuring that the result remained under the configured notebook directory. This could allow notebook files or directories to be moved, written, or deleted outside the notebook root. This issue affects Apache Zeppelin versions 0.9.0 through 0.12.0. Users are recommended to upgrade to version 0.12.1, which fixes this issue. |
| CVE-2026-46593 | A SQL injection vulnerability has been identified in the PHP Jabbers - PHP Poll Script. Improper neutralization of input provided by user to pjAdminPolls.controller.php endpoint allows an authenticated attacker to perform SQL Injection attacks.
This issue was fixed in version 4.1. |
| CVE-2026-46594 | A reflected cross-site scripting (XSS) vulnerability has been identified in the PHP Jabbers - PHP Poll Script. A malicious attacker can craft a specially crafted URL that, when opened, results in arbitrary JavaScript execution in the victim's browser.
This issue was fixed in version 4.1. |
| CVE-2026-50986 | PrestaShop module, totadministrativemandate <1.8.1 is vulnerable to Cross Site Request Forgery (CSRF). The payment validation controller has no CSRF token. An attacker can confirm an order in an awaiting status by hijacking a link. |
| CVE-2026-51785 | An issue in Hugo Leisink Hiawatha v.12.1 and before allows a remote attacker to execute arbitrary code via a crafted request |
| CVE-2026-51953 | An issue in FeehiCMS v.2.1.1 allows an attacker to escalate privileges via the Session management module, authentication logic, logout handler components |
| CVE-2026-52134 | An issue in the parseGoosePayload() function (/goose/goose_receiver.c) of libiec61850 v1.6 allows attackers to bypass authentication via a captured GOOSE frame. |
| CVE-2026-52371 | A Server-Side Request Forgery (SSRF) in the xxl-job-admin/jobinfo/trigger component of xxl-job v3.4.0 allows authenticated attackers to scan resources via supplying a crafted HTTP request. |
| CVE-2026-52855 | Wings is the server control plane for Pterodactyl, a free, open-source game server management panel. Prior to 1.12.3, {{config.}} placeholders in egg configuration-file templates allow a low-privileged user to read {{config.token}}, {{config.token_id}}, and {{config.docker.registries}} from the full daemon configuration. This issue is fixed in version 1.12.3. |
| CVE-2026-52856 | Wings is the server control plane for Pterodactyl, a free, open-source game server management panel. Prior to 1.13.0, a malformed packet received during the SFTP connection handshake causes a Go panic. This issue is fixed in version 1.13.0. |
| CVE-2026-52857 | Wings is the server control plane for Pterodactyl, a free, open-source game server management panel. Prior to 1.13.0, unbounded json, yaml, and xml configuration-file parsers in parser.go can process an oversized non-file parser configuration file and exhaust Wings process memory. This issue is fixed in version 1.13.0. |
| CVE-2026-53500 | Thumbor is an open-source photo thumbnail service by globo.com. Prior to 7.8.0, the ALLOWED_SOURCES configuration passes plain strings to re.match() without escaping dots, so a hostname differing at dot positions can match the allowlist. This issue is fixed in 7.8.0. |
| CVE-2026-53501 | Thumbor is an open-source photo thumbnail service by globo.com. Prior to 7.8.0, Thumbor’s HMAC validation can be bypassed due to the use of Python’s .replace() when removing the signature from the URL before validation. Since .replace() removes all occurrences of the substring, an attacker can insert the same signature multiple times in the URL and manipulate the final URL used for validation. This allows crafting URLs where the validated string differs from the actual requested resource, enabling loading images from unintended domains or paths. This issue is fixed in 7.8.0. |
| CVE-2026-53502 | Thumbor is an open-source photo thumbnail service by globo.com. Prior to 7.8.0, file_loader decodes percent-encoded path segments after its root-boundary validation, allowing traversal outside FILE_LOADER_ROOT_PATH through watermark or frame filter input. This issue is fixed in 7.8.0. |
| CVE-2026-53503 | Thumbor is an open-source photo thumbnail service by globo.com. Prior to 7.8.0, Thumbor's filters:convolution(<matrix>, <columns>, <should_normalize>) filter passes the user-controlled <columns> value to a C extension (thumbor/ext/filters/_convolution.c) where it is used as a divisor (for % and /) without validating columns > 0. When columns=0, the C code triggers undefined behavior; on x86_64 this reliably results in a fatal divide-by-zero trap (SIGFPE) and crashes the Thumbor process, causing a remote denial of service. This issue is fixed in 7.8.0. |
| CVE-2026-53504 | Thumbor is an open-source photo thumbnail service by globo.com. Prior to 7.8.0, the convolution filter regular expression performs exponential backtracking on crafted repeated numeric input, allowing a URL request to exhaust processing time. This issue is fixed in 7.8.0. |
| CVE-2026-53505 | Thumbor is an open-source photo thumbnail service by globo.com. Prior to 7.8.0, Thumbor's filters:proportion(<value>) filter does not enforce an upper bound on <value> and runs in the post-transform phase. An attacker can trigger extremely large resizes (CPU/memory exhaustion) and cause denial of service. This issue is fixed in 7.8.0. |
| CVE-2026-53510 | Savon is a Ruby SOAP client. From 0.9.8 until 2.17.2, Savon::Model .all_operations interpolates attacker-controlled WSDL operation names into Ruby source passed to module_eval, allowing Ruby code execution in the application process. This issue is fixed in version 2.17.2. |
| CVE-2026-53551 | free5GC is an open-source implementation of the 5G core network. Prior to 1.4.5, the free5GC AUSF (Authentication Server Function) does not validate the supiOrSuci field in UE authentication requests. Null bytes (\x00) and other control characters pass through JSON parsing unchanged and are forwarded to the UDM in an unescaped URL path. This causes Go's net/url.Parse() to fail, returning HTTP 500 "System failure" and leaking internal stack traces. An unauthenticated attacker can trigger this at scale causing denial of service for all subscribers attempting authentication through the affected AUSF. This vulnerability is fixed in 1.4.5. |
| CVE-2026-53599 | REDAXO is a PHP-based content management system. From 5.18.2 until 5.21.1, rex_mediapool::isAllowedExtension in redaxo/src/addons/mediapool/lib/mediapool.php lets an authenticated backend user with media[upload] permission upload a JPEG/PHP polyglot named shell.php.any.jpg, which web servers with multi-extension PHP handlers can execute as the web-server user. This issue is fixed in version 5.21.1. |
| CVE-2026-54785 | gemini-bridge is a lightweight MCP server bridging AI agents to Google's Gemini AI via the official CLI. From 1.0.0 until 1.3.1, consult_gemini_with_files in inline mode read any file path supplied in the files argument without confining it to the working directory, then forwarded the contents to the Gemini CLI. Because the caller also controls query, the file contents are echoed back through the Gemini round-trip (and sent to Google), making this an arbitrary local file read. This issue is fixed in version 1.3.1. |
| CVE-2026-54787 | sigstore-go is a Go library for Sigstore signing and verification. Prior to 1.2.1, sigstore-go does not check a bundle signing timestamp against the validity window of an ExpiringKey wrapping a self-managed long-lived signing key without a certificate, which can allow an attacker holding expired key material to sign accepted bundles. This issue is fixed in version 1.2.1. |
| CVE-2026-55100 | hashi-vault-js is a Node.js module for interacting with the HashiCorp Vault API. Prior to 0.5.2, src/Vault.js concatenates unencoded identifier values including name, username, group, role, and version into Vault request paths and query strings instead of using encodeURIComponent() and URLSearchParams, allowing path traversal and query parameter injection. This issue is fixed in version 0.5.2. |
| CVE-2026-55495 | Cloudreve is a self-hosted file management and sharing system. Prior to 4.17.0, the WOPI PUT_RELATIVE handler passes X-WOPI-SuggestedTarget to URI.JoinRaw as a path rather than a filename, allowing slash and dot-dot segments to escape the source file directory and create or conditionally overwrite files elsewhere in the same owner account. This issue is fixed in version 4.17.0. |
| CVE-2026-55496 | Cloudreve is a self-hosted file management and sharing system. Prior to 4.17.0, GET /api/v4/user/search calls SearchActive without adding a StatusActive predicate and serializes matches at RedactLevelUser, allowing any logged-in user to enumerate email addresses and profile metadata for inactive or banned accounts. The service calls userClient.SearchActive, but despite its name that method filters only by email/nickname keyword and never adds a StatusActive predicate — while the sibling lookups GetActiveByID and GetActiveByDavAccount, defined a few lines above it, do. Search hits are serialized at RedactLevelUser, which includes the email address. This issue is fixed in version 4.17.0. |
| CVE-2026-55497 | Cloudreve is a self-hosted file management and sharing system. Prior to 4.17.0, the built-in thumbnail and avatar image decoders limit compressed file size but do not limit decoded pixel dimensions, allowing an authenticated user to submit a small PNG, JPEG, or GIF that triggers an unbounded allocation and terminates the Cloudreve process through fatal out-of-memory behavior. This issue is fixed in version 4.17.0. |
| CVE-2026-55499 | Cloudreve is a self-hosted file management and sharing system. Prior to 4.17.0, a single-file share event-stream subscription resolves the share root to the owner’s parent folder and subscribes to that folder topic, allowing an authenticated share recipient to receive names, paths, rename targets, event types, and hashed identifiers for unshared sibling files and folders. This issue is fixed in version 4.17.0. |
| CVE-2026-55502 | Cloudreve is a self-hosted file management and sharing system. Prior to 4.17.0, POST /api/v4/admin/policy/oauth/signin requires only Admin.Read even though GetOauthRedirectService persists caller-supplied OneDrive secret and app_id values, allowing an OAuth token without Admin.Write to modify storage policy credentials. The route is inside the admin group that requires Admin.Read, but it does not add the local Admin.Write guard used by sibling policy mutation routes. Its handler persists attacker-supplied secret and app_id values into the selected OneDrive storage policy before returning an OAuth URL. This issue is fixed in version 4.17.0. |
| CVE-2026-55824 | Contao is an Open Source CMS. In versions 4.13.40 through 5.3.46 and 5.7.0-RC1 through 5.7.6, the crawler leaks auth credentials to external hosts. Contao's crawler tries to prevent confidential HTTP client options from being sent to external domains by creating a scoped client: full options for root page origins, cleaned options for everything else. The cleaner removes Cookie and Authorization headers, but it removes the non-Symfony option names basic_auth and bearer_auth instead of Symfony HttpClient's real auth_basic and auth_bearer options. When contao.crawl.default_http_client_options contains Basic or Bearer authentication for a protected staging/production site, those credentials remain in the "clean" client used for external links or configured additional URIs. An attacker who can get an external URL crawled, for example through a link on a crawled page while the broken-link checker is enabled, can receive the crawler credentials. This issue has been fixed in versions 5.3.47 and 5.7.7. |
| CVE-2026-55825 | Contao is an Open Source CMS. In versions 5.7.0 through 5.7.6, an authenticated backend user who can access one job can request an attachment identifier containing ../ segments and make the job attachment download endpoint read a file from another job directory inside var/job-attachments. The controller authorizes only the jobUuid route parameter. The later attachment lookup joins that authorized job UUID with the attacker-controlled identifier, then passes the combined path to the virtual filesystem. VirtualFilesystem::resolve() canonicalizes the whole path and only rejects paths that escape the filesystem mount, so authorized-job/../victim-job/debug_log.csv becomes victim-job/debug_log.csv. This is a cross-job authorization bypass for known job attachment paths. It is not a practical brute-force against unknown jobs because job directories are UUID v4 values. |
| CVE-2026-56670 | ComfyUI is a modular diffusion model GUI, api and backend with a graph/nodes interface. Prior to 0.28.0, the /view endpoint served uploaded SVG files inline because image/svg+xml and related XML content types were absent from the dangerous-content-type handling, allowing stored cross-site scripting in the ComfyUI origin. This issue is fixed in version 0.28.0. |
| CVE-2026-56671 | ComfyUI is a modular diffusion model GUI, api and backend with a graph/nodes interface. Prior to 0.28.0, get_model_preview in app/model_manager.py joins an unrestricted filename route capture to a selected model directory without a containment check, allowing an unauthenticated remote attacker to use traversal, encoded traversal, absolute paths, or an unbounded path_index to read image-decodable files and enumerate host paths. get_model_preview (app/model_manager.py) built the path with os.path.join(folder, filename) where filename is an unrestricted {filename:.*} route capture. Literal ../, percent-encoded %2e%2e%2f, and absolute paths all escaped the model directory; path_index was also unbounded. The target file is piped through Pillow and re-encoded as WEBP, so disclosure is limited to image-decodable files plus a file-existence/enumeration oracle (and internal-path leakage via path_index errors). This issue is fixed in version 0.28.0. |
| CVE-2026-56672 | ComfyUI is a node-based diffusion model GUI, API, and backend. Prior to 0.28.0, GET /userdata/{file} served user-controlled HTML and SVG files with extension-derived content types, allowing stored cross-site scripting in the ComfyUI origin and access to browser-stored API tokens, settings, workflows, and authenticated-equivalent API calls. The handler used web.FileResponse(path), so an uploaded .html/.svg was served as text/html/image/svg+xml. POST /userdata stores arbitrary request bodies (confined to the user's userdata directory). When a victim navigated to the file URL, the embedded script executed same-origin. The /view endpoint already forced dangerous MIME types to download; that protection had never been applied to /userdata. This issue is fixed in version 0.28.0. |
| CVE-2026-56673 | ComfyUI is a modular diffusion model GUI, API, and backend with a graph-and-node interface. Prior to 0.28.0, folder_paths.get_annotated_filepath and exists_annotated_filepath join workflow-controlled annotated filenames to a base directory without a containment check, allowing an unauthenticated crafted POST /prompt workflow using LoadImage or sibling nodes to probe arbitrary host paths and exfiltrate image-format files through /view. LoadImage defines a VALIDATE_INPUTS method, which causes the execution engine to skip COMBO (input-directory) validation. Affected nodes include LoadImage, LoadImageMask, LoadImageOutput, LoadAudio, LoadLatent, LoadVideo, and Load3D. This issue is fixed in version 0.28.0. |
| CVE-2026-57232 | Contao is an Open Source CMS. From 5.3.35 through 5.3.47 and from 5.7.0-RC1 through 5.7.8, the Feed Reader front-end module passes configured RSS feed URLs from FeedReaderController::getResponse() to feedIo->read() without scheme or private-address validation, allowing a backend user with module-edit permissions to make the server request internal network services, loopback addresses, or cloud metadata endpoints. In core-bundle/src/Controller/FrontendModule/FeedReaderController.php, the getResponse() function iterates over the configured feed URLs and passes each one directly to the HTTP client (via $this->feedIo->read($url, new Feed())) with no validation, while the DCA field definition for rss_feed in tl_module.php carries no URL scheme or host validation and the HTTP client is wired as @psr18.http_client (Symfony HttpClient) with no SSRF protection configured, since NoPrivateNetworkHttpClient is not used. This issue is fixed in versions 5.3.48. |
| CVE-2026-62323 | Cloudreve is a self-hosted file management and sharing system. Prior to 4.17.0, ViewerSessionValidation uses only the session-id prefix of a WOPI access token and does not enforce the requested viewer action, allowing a malicious or compromised WOPI viewer with a view session to forge the token suffix and invoke WOPI write routes for the underlying file. This issue is fixed in version 4.17.0. |
| CVE-2026-62324 | Jodit Editor is a WYSIWYG editor with a built-in file browser & image editor. Prior to 4.12.31, Jodit's sanitizeHTMLElement method fails to use isDangerousUrl to normalize javascript: href values before checking the scheme, allowing case variants, control-byte prefixes, and embedded tabs or newlines to bypass filtering and execute attacker-controlled script when a victim clicks a stored link rendered by an application. This issue is fixed in version 4.12.31. |
| CVE-2026-62959 | Coturn is a free open source implementation of TURN and STUN Server. From 4.5.2 through 4.14.0, when Coturn is started with --acme-redirect <URL> and exposes a plaintext-TCP listener, an unauthenticated remote client can send a single ordinary HTTP GET request and receive a 301 response whose Location header contains up to ~870 bytes of adjacent process heap memory. The leaked region is a recycled network receive buffer that is reused without being zeroed, so on a busy server it can contain data from other clients' requests (TURN credentials, OAuth tokens, relayed payloads). Root cause is a signed→unsigned conversion. This issue is fixed in version 4.15.0. |
| CVE-2026-62999 | Copier is a library and CLI app for rendering project templates. From 9.5.0 through 9.16.0, percent-encoded parent-directory segments or encoded path separators in a template URL can match a configured trusted repository prefix before an HTTP server or Git transport decodes the path, allowing unsafe template features from a repository outside the trusted prefix to run after user interaction. This issue is fixed in version 9.17.0. |
| CVE-2026-64607 | HttpClient based on the classic i/o model fails to correctly release the underlying connection back to the connection manager if it encounters an invalid or unsupported `Content-Encoding` header value in the response message. Please note this defect does not affect HttpClient based on the async i/o model.
This issue affects Apache HttpComponents Client: from 5.0-alpha1 through 5.6.2. |
| CVE-2026-65309 | ANDRITZ HIPASE-250 (formerly 250 SCALA) in affected versions stores
and transmits user passwords using a reversible format instead of a
one-way password hash. This allows an attacker able to read the
credential store or capture network traffic to recover all stored
passwords. |
| CVE-2026-65310 | ANDRITZ HIPASE-250 (formerly 250 SCALA), in the default configuration
of affected versions, exposes its data and configuration endpoint
without any authentication and permissive CORS on every response. An
unauthenticated attacker with network access can read live process
values and server configuration. |
| CVE-2026-65311 | The HTTP server component of ANDRITZ HIPASE-250 (formerly 250 SCALA)
in affected versions exposes an undocumented endpoint that changes
the server's logging level and target without requiring
authentication. A remote, unauthenticated attacker with network
access to the service may suppress audit logging, potentially
concealing other activity on the system. |
| CVE-2026-65313 | A provisioning script used when installing HIPASE-250 (formerly 250
SCALA) engineering workstations sets a fixed, hard-coded x11vnc
password. Because the same credential is applied to every workstation
provisioned this way, an attacker with adjacent-network access who
knows the password can gain VNC access to affected workstations. |
| CVE-2026-68770 | sentence-transformers contains a security control bypass vulnerability that allows attackers to achieve arbitrary code execution by exploiting a logic flaw in the import_module_class helper within sentence_transformers/util/misc.py, where the guard condition includes an 'or os.path.exists(model_name_or_path)' clause that satisfies the trust gate whenever the supplied path exists on the local filesystem, regardless of the trust_remote_code=False argument. Attackers who can control or influence the contents of a model directory on disk can place malicious Python files such as modeling_*.py referenced via modules.json, causing the code to execute at import time when an application loads the model with SentenceTransformer(path, trust_remote_code=False), bypassing the documented security contract and achieving code execution within the loading process. |
| CVE-2026-68771 | ComfyUI v0.23.0 contains an unsafe deserialization vulnerability in the LoadTrainingDataset node that allows unauthenticated remote attackers to execute arbitrary Python code by uploading a crafted pickle file and triggering its deserialization. Attackers can upload a malicious shard_*.pkl file via the unauthenticated POST /upload/image endpoint and then queue a workflow graph via POST /prompt referencing the uploaded file, causing torch.load to deserialize the attacker-controlled pickle payload using __reduce__ and execute arbitrary commands as the ComfyUI process user. |
| CVE-2025-71402 | better-auth versions greater than 1.3.34 and before 1.4.0 contain a vulnerability in the multi-session plugin's /sign-out after-hook, which trusts raw multi-session cookies and forwards extracted values to internalAdapter.deleteSessions without verifying the cookie signature (e.g., via getSignedCookie). An attacker can supply a forged _multi-* cookie to trigger deletion of arbitrary session tokens. |
| CVE-2025-71403 | better-auth versions before 1.1.20 contain a bypass vulnerability in trustedOrigins validation logic affecting absolute URLs and wildcard domains. Attackers can construct malicious callbackURL parameters that pass origin checks and trigger open redirects to steal sensitive tokens for account takeover. |
| CVE-2025-71404 | better-auth versions after v0.0.2 and before 1.1.16 contain a reflected cross-site scripting (XSS) vulnerability on the /api/auth/error page, where the value of the 'error' URL parameter is reflected as HTML without proper neutralization. An attacker who coerces a user into visiting a specially-crafted URL can execute arbitrary JavaScript in the context of the user's browser. The issue is fixed in version 1.1.16. |
| CVE-2026-10773 | The DHCPv4 client helper net_dhcpv4_msg_type_name() in subsys/net/lib/dhcpv4/dhcpv4.c indexes a static 8-element const char * name table after a faulty bounds check. The guard used msg_type <= sizeof(name) instead of msg_type <= ARRAY_SIZE(name); sizeof returns the byte size of the pointer array (32 on 32-bit, 64 on 64-bit targets) rather than the element count of 8, so message-type values from 9 up to that byte size pass the check and cause name[msg_type - 1] to read past the end of the array.
The msg_type value originates from the DHCP MESSAGE TYPE option, which is read as an unchecked raw byte from a received packet (net_pkt_read_u8) and passed unmodified into the lookup. A DHCP server, or any host able to inject a spoofed DHCP reply onto the client's link, can therefore drive the index out of bounds. The out-of-range slot yields a garbage const char * that is then dereferenced by a %s log conversion.
The lookup is reached only from a debug log statement (NET_DBG / LOG_DBG), so the out-of-bounds read is triggerable only when the DHCPv4 log module is built at DEBUG level (CONFIG_NET_DHCPV4_LOG_LEVEL_DBG), which is not the default configuration. When that condition holds, the result is an out-of-bounds read and a wild-pointer dereference: most likely a crash of the DHCP client (denial of service) and potentially disclosure of an adjacent pointer's contents through the log output. The fix replaces sizeof with ARRAY_SIZE, restoring the correct 1..8 acceptance window. |
| CVE-2026-18556 | Authentication bypass using an alternate path or channel vulnerability in N-able N-central allows Authentication Bypass.
This issue affects N-central: through 2026.1. |
| CVE-2026-2411 | Zephyr's Bluetooth host declares a GATT characteristic as two consecutive attributes: a Characteristic Declaration whose permission is hard-coded to BT_GATT_PERM_READ, and a Characteristic Value attribute that carries the application-specified security permissions (e.g. BT_GATT_PERM_READ_ENCRYPT / READ_AUTHEN / READ_LESC). The public notify and indicate APIs explicitly accept either attribute, and passing the declaration is the documented, common idiom. Before sending each notification or indication, the host re-checks link security with bt_gatt_check_perm() against params->attr in gatt_notify(), gatt_indicate(), and gatt_notify_multiple_verify_params() (subsys/bluetooth/host/gatt.c).
When the application passed the Characteristic Declaration attribute, the host correctly redirected the value handle but left params->attr pointing at the declaration, so the security check evaluated the declaration's permissions (no security required) instead of the value's. As a result the encryption/authentication/LESC requirement configured on the characteristic value was skipped. The Notify-Multiple path additionally used a mask that omitted the LE Secure Connections requirement.
A remote peer triggers the disclosure by connecting (optionally without pairing or encryption) and writing the Client Characteristic Configuration descriptor to enable notifications or indications, causing the server to emit the protected value over a link that has not reached the required security level. The impact is information disclosure / access-control bypass for characteristic values the application intended to expose only over a secured link; exposure depends on the application declaring encrypt/authen-required notify/indicate characteristics and on the CCC being writable at a lower security tier. There is no memory-safety or availability impact.
The fix adds bt_gatt_attr_resolve_value(), which maps a declaration attribute to the following value attribute before the permission check, and switches the Notify-Multiple path to the full BT_GATT_PERM_READ_ENCRYPT_MASK so the LESC requirement is also enforced. |
| CVE-2026-67307 | Wazuh 5.0.0-beta1 (fixed in 5.0.0-beta3) does not validate or override the cluster_name and cluster_node fields in inventory-sync Start FlatBuffer messages, while validating only the agentid against the authenticated agent identity. This allows a low-privileged enrolled agent to spoof cluster attribution in indexed inventory and vulnerability documents by forging wazuh.cluster.name values and influencing the document _id prefix, potentially tampering with inventory records or, in shared-indexer multi-cluster deployments, poisoning another cluster's records when numeric agent IDs collide. |
| CVE-2026-67308 | Wazuh workflows before 44bf114 contain a shell injection vulnerability in GitHub Actions that allows attackers to execute arbitrary commands by submitting pull requests with crafted VERSION.json files. Attackers can inject shell metacharacters into environment variables that are directly interpolated into run steps, enabling command execution and exfiltration of secrets including GITHUB_TOKEN and AWS credentials on self-hosted runners. |
| CVE-2026-67340 | ArcadeDB before 26.7.2 (arcadedb-engine) allows trigger scripts to look up host classes in java.lang.* (via Java.type) because ScriptTriggerExecutor adds java.lang.* to the allowed packages. An authenticated user with UPDATE_SCHEMA permission can create a JavaScript trigger that invokes java.lang.Runtime.getRuntime().exec() (or ProcessBuilder), achieving OS command execution when the trigger fires. |
| CVE-2026-67341 | ArcadeDB versions before 26.7.2 fail to enforce scripting authorization checks on the SQL DEFINE FUNCTION statement with LANGUAGE js. Attackers with database access can execute arbitrary JavaScript code by submitting DEFINE FUNCTION statements, bypassing security controls intended to restrict scripting to administrators. |
| CVE-2026-67342 | ArcadeDB versions before 26.7.2 contain an authorization bypass vulnerability in HTTP handlers for time series, batch, Prometheus, and Grafana endpoints that fail to validate database access permissions. Attackers can access and modify databases they are not authorized to use by directly calling affected endpoints with arbitrary database parameters. |
| CVE-2026-67343 | ArcadeDB versions before 26.7.2 fail to properly redact the cluster token in the GET /api/v1/server endpoint, allowing authenticated users to retrieve the arcadedb.ha.clusterToken value in cleartext. Attackers can use the leaked token with X-ArcadeDB-Cluster-Token and X-ArcadeDB-Forwarded-User headers to impersonate root and execute administrative actions including user creation, database operations, and server shutdown. |
| CVE-2026-67344 | ArcadeDB before 26.7.2 fails to enforce the UPDATE_SCHEMA database permission on the ALTER TYPE ... CUSTOM and ALTER TYPE ... BUCKETSELECTIONSTRATEGY SQL operations, which map to setCustomValue and setBucketSelectionStrategy in LocalDocumentType. An authenticated user with only read access (e.g., a read-only API token) can submit these ALTER TYPE statements via the HTTP command endpoint to mutate a type's custom schema metadata and bucket-selection strategy, bypassing the documented updateSchema permission boundary and potentially corrupting schema metadata and record routing. |
| CVE-2025-71400 | better-auth passkey versions before 1.4.0 contain an insecure direct object reference vulnerability in the passkey deletion endpoint that allows authenticated users to delete arbitrary passkeys by ID. Attackers with valid sessions can submit crafted requests to the delete-passkey endpoint with enumerated passkey IDs to remove other users' passkeys. |
| CVE-2025-71401 | better-auth (npm) before 1.4.2 allows an external request to configure baseURL when it is not otherwise defined (e.g., BETTER_AUTH_URL is unset). An attacker able to make the very first request to the server after startup can poison the router's base path, causing all routes to return 404 for all users (denial of service). The issue is not reachable when baseURL is explicitly configured or on typical managed hosting platforms. |
| CVE-2026-10774 | Zephyr's Bluetooth Mesh subnet key management leaks one PSA Crypto key slot on every subnet-key teardown. In subsys/bluetooth/mesh/subnet.c, net_keys_create() imports the Private Beacon Key into a PSA key slot under CONFIG_BT_MESH_PRIV_BEACONS (enabled by default), but subnet_keys_destroy() guarded the matching psa_destroy_key() with CONFIG_BT_MESH_V1d1. That Kconfig symbol was removed when explicit Mesh 1.0.1 support was dropped, so the destroy branch became permanently dead code and the import is never balanced by a destroy.
The imbalanced teardown is reached every time subnet keys are destroyed: deleting a subnet (Config Server NetKey Delete), completing a Key Refresh Procedure (which retires the old key set), and resetting/re-provisioning the node. The over-the-air triggers are processed only under the node's device key, so they are exercisable by the provisioner or network administrator that owns the node, reachable over the Bluetooth Mesh network.
With the default CONFIG_MBEDTLS_PSA_KEY_SLOT_COUNT of 16, repeated add/delete or key-refresh cycles exhaust the shared PSA key-slot pool after roughly a dozen rounds. Once exhausted, bt_mesh_private_beacon_key() and thus subnet creation fail: the node can no longer add subnets or complete key refresh, and other PSA crypto consumers on the device may be starved, until the device is rebooted. The fix aligns the destroy guard with the import guard (CONFIG_BT_MESH_PRIV_BEACONS) so each slot is freed. |
| CVE-2026-10848 | The OCPP 1.6 client in subsys/net/lib/ocpp parsed inbound WAMP RPC frames in parse_rpc_msg() (subsys/net/lib/ocpp/ocpp_j.c) using a hand-rolled helper, extract_string_field(), that copied the message's uid and action fields with strncpy(out_buf, token + 1, outlen - 1) and then scanned the result with strchr(out_buf, '"'). Because strncpy does not NUL-terminate the destination when the source is at least outlen - 1 (127) bytes long, the subsequent strchr reads past the 128-byte destination buffer into adjacent stack memory; if a " byte is found beyond the buffer, a one-byte out-of-bounds NUL write also occurs. A related defect in extract_payload() runs strchr/strrchr over the receive buffer, which may not be NUL-terminated when a maximal-length frame fills it.
The parsed bytes come directly from the OCPP central-system server over a websocket: the reader thread fills recv_buf via websocket_recv_msg() and calls parse_rpc_msg() on each inbound DATA frame (subsys/net/lib/ocpp/ocpp.c). A malicious or compromised central server, or an on-path attacker (OCPP is commonly deployed over plain ws://), can send an RPC frame whose uid or action field is 127+ bytes with no closing quote, triggering the out-of-bounds access.
The primary impact is a remotely triggerable denial of service: the unbounded scan can fault on an unmapped page, and the stray NUL write can corrupt adjacent stack state. The over-read data is not reflected to the peer, so disclosure is limited. The feature is EXPERIMENTAL and must be explicitly enabled (CONFIG_OCPP). The fix replaces the manual parser with the bounds-respecting json_mixed_arr_parse() and copies the extracted uid with an explicitly NUL-terminated buffer, eliminating both over-reads. |
| CVE-2026-14920 | ## Summary |
| CVE-2026-18570 | A flaw was found in the full-scope-disabled client-policy executor within the keycloak-services component. This component is responsible for enforcing security policies during client registration and configuration in Red Hat Build of Keycloak. The issue occurs because the executor only validates the fullScopeAllowed field when it is explicitly provided in a request. By omitting this field, a delegated user can bypass the policy, resulting in a client created with full scope access. This allows the client to obtain tokens with unauthorized role mappings. |
| CVE-2026-18571 | A flaw was found in the user creation component of Keycloak when Fine-Grained Admin Permissions V2 (FGAP V2) is enabled. This issue allows a sub-administrator with permission to create users to add those users to any group, even groups the sub-administrator is not authorized to manage. This could lead to unauthorized access to sensitive information or elevated privileges for the newly created users. |
| CVE-2026-18572 | Keycloak provides authorization services that allow administrators to restrict access to resources based on time policies (for example, only allowing access during business hours). A flaw was discovered where a user can include a fake time value in their authorization request that overrides the actual server time. This allows the user to bypass these time-based restrictions and access protected resources at unauthorized times. |
| CVE-2026-18573 | A flaw was found in the keycloak-services component of Keycloak, which is used for managing authentication and authorization flows. The issue occurs when a realm administrator configures client policies to enforce specific authentication requirements on confidential clients. Due to improper evaluation of the client state during an update operation, an attacker with client management permissions can bypass these security policies by first creating a public client and then updating it to a confidential client with weaker authentication. This can result in the persistence of clients that do not comply with the intended security hardening of the realm. |
| CVE-2026-18577 | An incomplete patch for CVE-2026-18556 allows for authentication bypass and account takeover in N-central Versions through 2026.3.1 |
| CVE-2026-3245 | A deserialization vulnerability in PRISMAproduction Version 6.5 or earlier that may lead to arbitrary code execution. |
| CVE-2026-65321 | PyAthena prior to 3.35.4 contains a sql injection vulnerability that allows unauthenticated attackers to inject arbitrary SQL by exploiting improper quote-escaping in DefaultParameterFormatter.format(), which routes DELETE and CTAS statements to the _escape_hive function that backslash-escapes single quotes rather than doubling them. Because Athena and Trino do not treat backslashes as escape characters inside string literals, attacker-supplied input such as a single quote followed by SQL syntax causes the parser to terminate the string literal prematurely, enabling data exfiltration via UNION SELECT, execution of destructive statements, and attacker-controlled CTAS destination and content. |
| CVE-2026-68581 | Vikunja versions 0.22.0 through 2.3.0 fail to validate the principal type in API token management. Because user IDs and link-share IDs are independent numeric sequences and both resolve through a generic web.Auth.GetID() interface, a link-share JWT whose numeric ID equals a target user's ID is treated as that user by the /api/v1/tokens endpoints. An authenticated attacker can obtain a target's numeric user ID via authenticated user search, then create link shares on an attacker-writable project until the link-share sequence reaches that value, and use the resulting link-share JWT to list, create, and delete the target user's API tokens (including issuing a new token with attacker-chosen scopes under the target's permissions). Fixed in version 2.4.0. |
| CVE-2026-68582 | Vikunja versions >= 0.24.0 and <= 2.3.0 contain a broken object level authorization (BOLA) vulnerability in the task-collection endpoint (GET /api/v1/projects/{project}/views/{view}/tasks). The endpoint loads the requested project view from the URL path without verifying the caller is authorized for it. For a link-share token holder, the task scope is pinned to the share's own project, but the view is taken from the attacker-controlled path and never re-validated. As a result, a holder of any project share link can read any other tenant's kanban bucket records — bucket titles and the full created_by user object (username, name, id) — for every view in the instance. The same missing pre-authorization view load also creates a project/view-ID existence oracle (404 vs. non-404) usable by link shares and ordinary authenticated users. Task contents remain constrained to the share's own project and are not disclosed. Fixed in 2.4.0. |
| CVE-2026-68583 | luci-app-adblock-fast before 1.2.4-4 contains a stored cross-site scripting vulnerability in the blocklist name field that allows lower-privileged users to inject active HTML. When an administrator views the AdBlock Fast status page, the injected payload executes in the administrator's browser under the LuCI origin. |
| CVE-2026-9335 | A vulnerability in keras-team/keras versions <= 3.14.0 allows arbitrary local HDF5 file content disclosure due to improper handling of HDF5 ExternalLinks. The `KerasFileEditor` and `keras.saving.load_weights` functions bypass the `safe_get_h5_group` and `safe_get_h5_dataset` helpers, which are designed to reject ExternalLinks and SoftLinks. This results in automatic dereferencing of links to external HDF5 files, enabling attackers to disclose sensitive data from the victim's local filesystem. Specifically, `KerasFileEditor` extracts attributes and datasets from linked files into its internal structures, while `keras.saving.load_weights` loads weights from linked files into the user's model. This issue can be exploited by providing a malicious `.h5`, `.weights.h5`, or `.keras` file containing ExternalLinks. |
| CVE-2026-9856 | A vulnerability in huggingface/transformers versions <=5.8.0.dev0 allows an attacker to perform arbitrary file writes via path traversal. The issue resides in the `save_pretrained()` methods of `PreTrainedTokenizerBase` and `ProcessorMixin`, where keys from the `chat_template` dictionary are used directly as filenames without proper validation. An attacker can exploit this by publishing a malicious Hugging Face Hub repository with a crafted `tokenizer_config.json` file. When a victim downloads and saves the tokenizer or processor, the attacker-controlled keys can escape the intended save directory, enabling arbitrary file writes with attacker-controlled content. This vulnerability affects multiple processors inheriting from `ProcessorMixin`, including Idefics, Florence, Gemma, Phi, and Qwen-VL. |