actix-http versions from 3.18.10 up to but not including 3.20.2 contain an HTTP request smuggling vulnerability in the chunked decoder. The flaw arises because the decoder incorrectly treats a bare LF after the last chunk as the start of a trailer section. This allows attackers to craft a request body ending with 0\r\n\n that can cause the next request's body to be parsed as new requests when passing through a connection-reusing proxy, potentially bypassing WAF or upstream protections.
actix-files versions before 0.7.4 contain an information disclosure vulnerability that allows remote unauthenticated attackers to read hidden files by exploiting Windows 8.3 short-name aliases. This bypasses the leading-dot check designed to protect sensitive files such as .env and .git directories, potentially exposing secrets and configuration data.
Rclone versions from 1.56.0 up to but not including 1.76.0 have a timing side-channel vulnerability in the htpasswd authentication middleware. This flaw allows unauthenticated attackers to enumerate valid usernames by sending HTTP Basic Auth requests with incorrect passwords and measuring response times. The vulnerability arises because bcrypt verification is only performed for existing users, causing observable timing discrepancies. The issue is fixed in version 1.76.0.
CVE-2026-134349 is a memory exhaustion vulnerability in rclone versions 1.46.0 up to but not including 1.76.0. The flaw exists in the remote control server where unauthenticated clients can send oversized JSON payloads to protected RC methods. The server decodes these payloads before performing authorization checks, which can lead to out-of-memory termination and disrupt ongoing transfers.
rclone versions 1.67.0 through 1.75.x contain a vulnerability in the Ulozto backend where authentication tokens are leaked to unintended hosts during HTTP redirects. This occurs because the downloads use an authenticated client instead of a CDN client, exposing long-lived API keys and session tokens. An attacker able to observe or redirect CDN download traffic could capture these tokens and gain persistent access to the victim's Ulozto storage.
Rclone versions from 1.49.0 up to but not including 1.76.0 have an authentication vulnerability where revoked users can retain access by reconnecting within five minutes due to cached authentication responses not being revalidated. This affects the serve sftp, ftp, http, and webdav features. The issue is fixed in version 1.76.0.
Rclone versions from 1.51.0 up to but not including 1.76.0 contain an authentication bypass vulnerability in the serve sftp --auth-proxy feature. The vulnerability arises because the cache key used for authentication does not include the authentication method, allowing attackers who know a victim's base64 public key to reuse cached authentication entries from the same client IP and gain unauthorized read-write access to the virtual file system (VFS). This flaw has been fixed in version 1.76.0.
rclone versions prior to 1.76.0 have a vulnerability where credentials are leaked in the swift, netstorage, and seafile backends due to insecure handling of HTTP redirects. This can expose authentication tokens or passwords to unauthorized actors if they intercept or observe redirects from the configured endpoints.
CVE-2026-134354 is a vulnerability in rclone versions 1.59.0 up to but not including 1.76.0. It involves improper link resolution in the local backend's setXattr function, allowing attackers to escape confinement by writing user.* extended attributes outside the intended destination root via a maliciously planted symlink. This can be exploited using the --links and --metadata options to set attacker-controlled extended attributes on files outside the target directory.
CVE-2026-134355 is an improper authorization vulnerability in AzuraCast versions before 0.24.1. It affects the /api/station/{station_id}/profile/edit endpoint, allowing authenticated station users with Manage Station Profile permission—but without Manage Station Broadcasting permission—to access and modify sensitive Icecast/Shoutcast passwords. This includes admin, source, relay, and streamer passwords. Exploitation can lead to unauthorized reading or overwriting of these passwords, potentially locking out legitimate operators.
AzuraCast versions from 0.19.0 up to but not including 0.24.1 contain an authorization bypass vulnerability. This flaw allows users with station media rights to add tracks to playlists of other stations by manipulating playlist IDs. Exploitation can result in unauthorized injection of entries into other stations' AutoDJ rotations and public now-playing metadata, potentially causing playback failures. The issue is fixed in version 0.24.1.
CVE-2026-134359 is a vulnerability in AzuraCast versions before 0.24.1 where improper escaping of backslashes in Liquidsoap annotation values allows attackers to inject process: protocol URIs. This can lead to OS command execution within the Liquidsoap container during media playback. The issue is fixed in version 0.24.1.
CVE-2026-134360 is a vulnerability in heymrun's heym product before version 0.0.148. It allows a signed-in user with workflow execution permissions to navigate to file:// URLs without validation, enabling reading of sensitive files such as Kubernetes ServiceAccount tokens or Docker secrets. The issue is fixed in version 0.0.148.
Coder versions 2.6.0 before 2.29.21, 2.30.0 before 2.34.10, 2.35.0 before 2.35.8, 2.36.0 before 2.36.5, and 2.37.0 before 2.37.1 contain a denial of service vulnerability in the workspace agent dRPC API. This occurs because certain handlers dereference unset request sub-messages and unconfigured Tailnet providers without recovering from panics. An attacker with a workspace agent token can send specific malformed messages to repeatedly crash the coderd service.
Coder versions 2.6.0 before 2.29.26, 2.30.0 before 2.34.18, 2.35.0 before 2.37.10, and 2.38.0 before 2.38.6 contain a cross-site request forgery (CSRF) vulnerability. This occurs because CSRF exemption patterns are matched as unanchored regular expressions, allowing certain API routes to bypass CSRF token and same-origin checks. An attacker controlling a same-site page can trick signed-in users' browsers into sending forged POST requests that may recreate development containers or add users with specific names to organizations. The vulnerability has a low CVSS score of 2.3 and has been fixed in versions 2.29.26, 2.34.18, 2.37.10, 2.38.6, and 2.37.0.
Coder versions 2.32.0 before 2.34.17, 2.35.0 before 2.36.0, 2.36.0 before 2.36.12, 2.37.0 before 2.37.9, and 2.38.0 before 2.38.5 contain a vulnerability where the read_skill_file tool improperly follows symbolic links outside the intended skill directory. This allows attackers to craft malicious skills with symlinked files that can leak other workspace files into the model context and chat history when executed.
pypdf versions from 3.7.1 up to but not including 7.0.0 contain an information disclosure vulnerability. This flaw allows attackers to infer placeholders in encrypted templated PDF documents by analyzing file identifiers. If an attacker has access to a document sharing the same template and can observe file identifiers from other documents, they may recover sensitive placeholder contents from encrypted PDFs. The vulnerability is rated high severity with a CVSS score of 8.7. A fix is available in version 7.0.0.
pypdf versions before 7.0.0 have a resource exhaustion vulnerability due to inefficient algorithmic complexity in handling PDF cross-reference streams. An attacker can craft a PDF with cross-reference stream offsets all pointing to a single large object, causing excessive memory consumption during parsing.
pypdf versions from 2.4.2 up to but not including 7.0.0 have a resource exhaustion vulnerability in their handling of XFA forms. An attacker can craft a PDF with many small XFA form fields that cause excessive memory consumption when loaded, leading to denial of service. This vulnerability is fixed in version 7.0.0.
CVE-2026-134370 is a path traversal vulnerability in the Frappe Framework affecting versions before 15.120.1 and 16.x before 16.33.1. It allows authenticated users with minimal privileges to read arbitrary files on the server outside the intended application boundaries via the file attachment handling functionality.