CVE mới & PoC public

1105 CVE

• CVE-2026-110354 Cao 7.5 CWE-835 22/10 10:16

  HAProxy 3 before 3.4.6 allows a remote unauthenticated attacker to cause a denial of service (persistent worker-thread CPU exhaustion) via an HTTP/1.1 chunked request. Although the H1 multiplexer drains the request before closing the stream, the parser repeatedly re-processes the same bytes without making progress. The affected thread stays busy at 100% CPU after timeouts expire and after the client disconnects, and repeated requests can consume all worker threads until the process is restarted.

• CVE-2026-110353 Cao 7.5 CWE-252 22/10 10:16

  HAProxy 3.2 before 3.4.6 allows a remote unauthenticated attacker to cause a denial of service (worker-thread CPU exhaustion) via a truncated HTTP/2 PING frame. h2c_handle_ping in src/mux_h2.c does not check for a short read, unlike the other H2 frame handlers. Because the incomplete frame cannot be processed, its data is never consumed, and the H2 multiplexer tasklet is repeatedly woken by that pending data without making progress. The CPU spins until timeout http-request closes the connection, and an attacker opening many such connections can sustain high CPU usage.

• CVE-2026-114727 Cao 8.3 CWE-829 22/10 09:17

  Roundcube Webmail before 1.6.20 and 1.7.x before 1.7.5 passes potentially unsafe content into ImageMagick when resizing images. SVG and EPS were accepted as valid image types for contact photo upload and conversion, and both formats can reference external resources and invoke ImageMagick delegates when processed. Upstream fixed this in the commit “Fix passing potentally unsafe attachment content into ImageMagick on image resize”, which removes svg and eps from the accepted types.

• CVE-2026-114724 Cao 7.4 CWE-502 22/10 09:17

  Roundcube Webmail before 1.6.20 and 1.7.x before 1.7.5 contains an object injection primitive in rcube_session::unserialize(). A session key prefixed with “!” was treated as having no value, and the reconstructed serialized string was emitted with a key but no corresponding value, desynchronizing the remainder of the input so that following session data was interpreted as a serialized value. This yields a PHP object injection primitive for an attacker able to influence stored session data. Upstream fixed this in the commit “Fix object injection primitive in rcube_session::unserialize() via “!”-prefixed session keys”, which rejects such keys.

• CVE-2026-114720 Cao 7.1 CWE-93 22/10 09:16

  Roundcube Webmail before 1.6.20 and 1.7.x before 1.7.5 is affected by an IMAP command injection vulnerability. A remote authenticated attacker can exploit this via CRLF sequences injected into the _flag parameter of the mail/mark action to execute arbitrary IMAP commands. Upstream fixed this in the commit “Fix IMAP command injection via CRLF in `_flag` parameter of mail/mark action”, which rejects flags containing control, whitespace and IMAP-special characters.

• CVE-2026-114716 Cao 8.5 CWE-99 22/10 09:16

  Roundcube Webmail before 1.6.20 and 1.7.x before 1.7.5 is affected by a mail header injection vulnerability. rcmail_sendmail::extract_inline_images() used the media type taken from an inline image data URI without validating it, so a remote authenticated attacker could inject arbitrary MIME part headers into the outgoing message by placing CRLF sequences in that media type. Upstream fixed this in the commit “Fix mail header injection via media type in an inline image data URI”, which validates the media type with rcube_mime::is_mimetype_valid().

• CVE-2026-114664 Cao 7.1 CWE-89 22/10 09:16

  Frappe Framework before 16.33.0 contains a blind SQL injection vulnerability in the frappe.desk.doctype.event.event.get_events method, which fails to properly neutralize user-supplied input before including it in SQL queries. Authenticated attackers can send crafted input to get_events and infer query results from application behavior to extract sensitive database information.

• CVE-2026-114663 Cao 8.6 CWE-706 22/10 09:16

  Vyper 0.4.0b1 through 0.5.0b1 miscompiles unfoldable constants and omitted default arguments from imported modules, because parse_Name resolves their names in the using function's scope rather than at declaration. When a contract's parameter, local, or loop variable shares such a name, attacker-influenced input can replace the intended value, including out-of-range values for narrow types such as uint8, corrupting validation or accounting logic.

• CVE-2026-114662 Cao 8.6 CWE-352 22/10 09:16

  Obot before 0.26.0 contains an authentication bypass vulnerability in the CLI login flow that allows unauthenticated attackers to steal API keys by choosing token request IDs. Attackers can create a token request via POST /api/token-request, send the login link to a signed-in user, and fetch the resulting non-expiring API key to take over the victim's account.

• CVE-2026-114660 Cao 7.0 CWE-470 22/10 09:16

  Grav from 2.0.0 before 2.2.5 contains an improper access control vulnerability in FlexDirectory::dynamicFlexField() that allows page authors to call arbitrary Flex object methods via flex-*@ directives in page frontmatter. When a user with page edit rights opens the page, directives calling setStorageKey() and delete() run in their session, deleting protected pages neither account may delete.

• CVE-2026-114659 Cao 7.1 CWE-424 22/10 09:16

  Grav Flex Objects plugin 1.4.0 before 1.4.17 contains an authorization bypass because its generic /api/v1/flex-objects/pages API routes check only directory-level admin.pages permissions, without enforcing the Pages API's api.pages.read/write permissions or page-level permissions rules. Authenticated non-super users with api.access and admin.pages or admin.configuration.pages permissions can read page source and frontmatter, modify content, or delete pages they are denied.

• CVE-2026-114658 Cao 8.7 CWE-269 22/10 09:16

  Grav Login plugin from 3.6.0 before 3.9.13 contains a missing authorization vulnerability in the login.invite task that allows page authors to obtain super user accounts. Attackers can define an invitation form granting super access in page frontmatter, submit it unauthenticated, and register the invited address for full site control.

• CVE-2026-114642 Cao 7.0 CWE-295 22/10 09:16

  Traefik 2.11.0 before 2.11.58 and 3.0.0 before 3.7.14 stores its NTLM/Kerberos backend round tripper in a client-connection slot not keyed by ServersTransport, so one service's TLS client configuration is reused for another service's backend. An unauthenticated client can arm the slot via a backend issuing an NTLM or Negotiate challenge, then reach another service on the same keep-alive connection, bypassing its root CAs, server name, and certificate verification.

• CVE-2026-114640 Cao 7.0 CWE-290 22/10 09:16

  Traefik from v3.7.0 before v3.7.14 contains an authentication bypass vulnerability in its Kubernetes Ingress-NGINX provider that forwards client-supplied Ssl-Client-* headers on the plaintext HTTP router. When ssl-redirect is disabled alongside auth-tls-pass-certificate-to-upstream, unauthenticated attackers can send Ssl-Client-Verify: SUCCESS with a chosen certificate subject and issuer to impersonate mTLS identities.

• CVE-2026-114639 Cao 7.0 CWE-287 22/10 09:16

  Traefik 2.11.0 before 2.11.58 and 3.0.0 before 3.7.14 contains an authentication bypass because a first request preauthenticated with an Authorization: Negotiate token authenticates a socket in the globally shared backend connection pool, which is never isolated. Unauthenticated attackers on separate frontend connections can be assigned that NTLM/Kerberos-bound socket and perform reads or writes with the victim's identity.