CVE-2026-134436 is a missing authorization vulnerability in Frappe ERPNext versions 15.8.0 before 15.121.4 and 16.x before 16.36.0. The flaw exists in an endpoint shared across modules that returns transaction child table rows without verifying permissions on the child or parent documents. Authenticated attackers, including portal users or users without roles, can exploit this to read business document line details by supplying document types and filter conditions.
Finsys Dockhand versions from 1.0.7 up to but not including 1.0.57 contain a path traversal vulnerability. This flaw allows users with stacks:edit permission to read or overwrite arbitrary files by setting an unvalidated envPath on a stack source. Exploitation involves using GET or PUT requests on the /api/stacks/[name]/env/raw endpoint to access sensitive files such as /etc/shadow or SSH keys. The vulnerability is fixed in version 1.0.57.
Dockhand versions before 1.0.57 have a cross-site request forgery (CSRF) vulnerability due to missing Origin header validation and absent CSRF tokens. This flaw allows attackers to trick authenticated users into submitting forged POST requests that can stop, start, or delete stacks, modify settings, and change credentials.
Dockhand versions before 1.0.57 have a clickjacking vulnerability due to missing X-Frame-Options or Content Security Policy frame-ancestors headers. This allows attackers on the same-site origin to embed the UI in an invisible iframe and trick authenticated users into performing unintended actions such as stopping stacks, deleting resources, or changing settings. The vulnerability is fixed in version 1.0.57.
Dockhand versions before 1.0.57 have a privilege escalation vulnerability that allows non-admin users with settings:edit permission to modify the OIDC provider configuration improperly. By sending a crafted PUT request to the OIDC API endpoint, attackers can set the adminClaim to their own OIDC group, resulting in full admin access upon their next OIDC login. This vulnerability is fixed in version 1.0.57.
CVE-2026-134452 is a high-severity vulnerability in Finsys dockhand before version 1.0.57. It involves an untrusted search path issue where stack secret environment variables are merged into the environment used to spawn docker compose during stack deployments without filtering reserved keys like PATH. An authenticated attacker with stacks:edit permission can exploit this by setting PATH to a directory containing malicious binaries, leading to arbitrary command execution as the Dockhand server process user.
Dockhand versions before 1.0.50 have an improper access control vulnerability allowing authenticated unprivileged users to access restricted API endpoints without proper RBAC checks. This flaw permits attackers to retrieve sensitive information such as user emails, roles, license keys, and environment details. Additionally, if the license is invalid, all users are granted admin rights. The vulnerability is rated high severity with a CVSS score of 7.1 and was fixed in version 1.0.50.
CVE-2026-134454 is an access control vulnerability in HyNetworks hysteria versions 2.0.0 up to but not including 2.9.2. It allows authenticated clients to bypass UDP destination blocking by reusing an established UDP session, enabling them to send packets to localhost or private network addresses and potentially access internal UDP services. The issue is fixed in version 2.9.2.
CVE-2026-134455 is a denial of service vulnerability in HyNetworks hysteria versions 1.0.0 up to but not including 2.9.2. The flaw occurs in the FragUDPMessage component, where UDP fragmentation does not properly handle headers larger than the maximum datagram payload size. Authenticated clients can exploit this by advertising a small max_datagram_frame_size, triggering a UDP response that causes a slice bounds panic and crashes the server. The issue is fixed in version 2.9.2.
Hysteria versions 2.5.0 up to but not including 2.9.2 have a denial of service vulnerability due to the HTTP sniff implementation not limiting header size. Attackers can send excessively large HTTP headers continuously, causing server memory exhaustion and crashes. This issue is fixed in version 2.9.2.
Hysteria versions from 2.5.0 up to but not including 2.8.2 contain a memory allocation vulnerability triggered when sniffing is enabled. Authenticated users can exploit this by sending a crafted QUIC packet with an excessively large crypto length, causing the server to exhaust memory and crash. This issue is fixed in version 2.8.2.
A cryptographic vulnerability exists in web-token jwt-framework and jwt-library versions before 3.4.12, 4.0.8, 4.1.10, and 4.2.4, as well as jwt-encryption-algorithm-ecdh-es through 3.2.10. The ConcatKDF::generate() function performs only a single SHA-256 round, limiting key derivation to 32 bytes for ECDH-ES and ECDH-SS key agreements. This flaw allows decryption of JWE tokens using A256CBC-HS512 with an all-zero AES key, and reduces the strength of A192CBC-HS384 content keys to 64 bits.
CVE-2026-134459 is a high-severity vulnerability in the web-token jwt-library affecting versions before 3.4.10, 4.0.7, and 4.1.7. It involves a padding oracle in the RSACrypt::decryptWithRSA15() function due to timing differences in RSA1_5 decryption. Remote attackers can exploit this by submitting crafted JWE encrypted_key values to recover the wrapped content encryption key.
A vulnerability in web-token jwt-framework and jwt-experimental prior to versions 3.4.10, 4.0.7, and 4.1.7 allows attackers to modify the encrypted content encryption key without detection due to a missing integrity check in the experimental Chacha20Poly1305 key-encryption algorithm. This flaw discards the Poly1305 authentication tag during encryption and omits it during decryption, weakening the cryptographic protection of the key.
CVE-2026-134461 is a high-severity vulnerability in the web-token jwt-library affecting versions before 3.4.10, 4.0.7, and 4.1.7. It involves uncontrolled resource consumption in the PBES2AESKW::unwrapKey() function, where an attacker can specify an excessively large iteration count in the p2c JWE header, causing lengthy PBKDF2 computations that exhaust server CPU resources. This vulnerability allows unauthenticated attackers to perform denial-of-service attacks by sending specially crafted JSON Web Encryption (JWE) tokens.
CVE-2026-133985 is a high-severity vulnerability in TRENDnet TEW-657BRM version 1.00.1 involving an out-of-bounds read in the /usr/sbin/restore_config.cgi file. The flaw can be exploited remotely without user interaction and requires low attack complexity and privileges. The product is end-of-life and no official patch or fix is available. The vendor has not confirmed the vulnerability but plans to notify customers. Exploit code has been published.
Red Hat has issued a critical security advisory for PHP 7.4 on Red Hat Enterprise Linux 8, addressing multiple vulnerabilities including out-of-bounds read/write, use-after-free, remote code execution, local privilege escalation, request smuggling, cross-site scripting, signature bypass, information disclosure, and denial of service. These vulnerabilities affect various PHP components such as PHP-CGI, PHP-FPM, PHAR, SOAP, and session handling. The update backports security fixes from PHP 8.2.36 to 7.4 and is available for multiple Red Hat Enterprise Linux 8 architectures.
A vulnerability in eigent-ai eigent versions up to 1.0.2 allows remote attackers to bypass authentication in the Local Agent Control API component. This issue affects an unknown functionality in the chat_controller.py file. The vulnerability is fixed in version 1.0.3.
CVE-2025-61963 is a high-severity vulnerability in the AmdPlatformRasSmm driver caused by improper input validation. An attacker with privileged access could exploit this flaw to overwrite SMRAM, potentially leading to arbitrary code execution. No patch information is currently available, and no known exploits are reported in the wild.
CVE-2026-28235 is a vulnerability in the AMD RAID Configuration Driver where improper address validation could allow an attacker with administrative privileges to read arbitrary kernel memory. This may lead to loss of confidentiality of sensitive information stored in kernel memory. The vulnerability requires administrative access and does not involve user interaction.