HeyForm: Uploaded HTML files are served as text/html on the app origin with no CSP — stored XSS

HeyForm is an open-source form builder. Prior to 3.0.1, /api/upload accepts a file according to the client-controlled multipart file.mimetype while retaining an HTML filename, and static/upload serves the resulting file as text/html with contentSecurityPolicy: false. An ordinary user can therefore upload active HTML that executes on the HeyForm origin when a victim opens its returned URL. The same-origin script can access /graphql with the victim's session and read or modify data available to that account. The code also permits upload requests tied to an active public form through formId, openToken, and a fieldId belonging to a file or signature question, although the advisory did not fully verify that unauthenticated path end to end. This issue is fixed in version 3.0.1.

fast-agent: `fast-agent serve` binds HTTP to 0.0.0.0 with no authentication by default; with --shell this is unauthenticated remote command execution

fast-agent is a framework for building, running, and evaluating AI agents with Model Context Protocol support. Prior to 0.7.21, fast-agent serve uses HTTP transport bound to 0.0.0.0 on port 8000 without authentication, allowing any network client that can reach the /mcp endpoint to invoke the exposed agent tool. When the server is started with --shell, the agent can call the local execute tool implemented by tools/local_shell_executor.py, so an unauthenticated remote client can cause operating-system commands to run as the server user; without --shell, the client can still consume the operator's model and configured MCP tools. The defaults are defined in cli/commands/serve.py and mcp/server/agent_server.py, and the non-interactive server permission prompt does not prevent the described access. This issue is fixed in version 0.7.21.

JupyterLab Desktop: Cross-Site Scripting (XSS) leading to Remote Code Execution on Connect URL

JupyterLab Desktop is the cross-platform desktop application for JupyterLab. Prior to 4.6.2-1, remote-server URLs were inserted into the server-selection dialog and Welcome view without safe encoding. An attacker who convinces a user to submit a crafted URL can execute JavaScript in the Welcome renderer and abuse its privileged IPC bridge to execute arbitrary code with the user’s privileges. The Welcome view also unsafely renders RSS news-feed titles; exploitation through this separate input path requires control over content delivered through the application’s Jupyter Blog feed or its cached feed data, rather than submission of a remote-server URL. This issue affects all supported platforms and is fixed in version 4.6.2-1.

Kroki Mermaid companion SSRF via flowchart image URLs

Kroki creates diagrams from textual descriptions. From 0.26.0 until 0.31.2, the optional Mermaid companion's mermaid/src/worker.js renders attacker-controlled flowchart image nodes in headless Chromium without a network request policy, even when KROKI_SAFE_MODE is set to secure. An actor who can submit Mermaid source and satisfy any authentication or authorization configured by the deployment can cause the companion to issue HTTP or HTTPS GET requests when requesting PNG output. Exploitation requires the Mermaid companion, PNG rendering through its headless Chromium process, a diagram-submission path available to the actor, and network reachability from the companion to the target. The demonstrated impact is limited to blind request reachability and possible state-changing GET endpoints; response exfiltration, cloud metadata access, port scanning, and access beyond the companion's network reachability are not established. This issue is fixed in version 0.31.2.

Modcluster-core: mod_cluster advertise listener: unauthenticated dos via crafted multicast datagram

A flaw was found in mod_cluster's AdvertiseListenerImpl (org.jboss.modcluster core module). A single crafted UDP multicast datagram with a valid HTTP status line and a "Server:" header but without the "Date:", "Digest:", and "Sequence:" headers triggers a NullPointerException in verifyDigest() that is not caught by the worker thread's exception handler. This causes the advertise listener thread to terminate permanently. The failure is silent (isListening() continues to return true) and persists until the node is restarted. The crash occurs before the AdvertiseSecurityKey comparison, so deployments with a configured security key are still affected.

In getSelectionBySubIds of ProviderUtil.java, there is a possible cross-user data leak due to a logic error in the code. This could lead to local information disclosure with no additional execution privileges needed. User interaction is not needed for exploitation.

Authenticated Local Arbitrary File Write Vulnerability in HPE Networking Fabric Composer

An arbitrary file write vulnerability exists in the underlying operating system of HPE Networking Fabric Composer. Successful exploitation could allow an authenticated local attacker to create or overwrite arbitrary files on the affected system with elevated privileges.

Privilege Escalation Vulnerability in HPE Networking Fabric Composer API

A privilege escalation vulnerability exists in the API of HPE Networking Fabric Composer if certain preconditions outside of the attacker's control are met. Successful exploitation could allow an authenticated attacker with low privileges to gain elevated privileges on the affected system.

chromium-browser: Uninitialized resource in WebGL

Uninitialized resource in WebGL in Google Chrome prior to 157.0.8103.38 allowed a remote attacker to read memory outside the sandbox via a crafted HTML page. (Chromium security severity: High)

chromium-browser: Out of bounds read in Animation

Out of bounds read in Animation in Google Chrome prior to 157.0.8103.38 allowed a remote attacker to potentially read memory inside the sandbox via a crafted HTML page. (Chromium security severity: Low)

CGI::Carp versions before 5.00 for Perl leak warnings across requests in persistent environments via the warningsToBrowser buffer

CGI::Carp versions before 5.00 for Perl leak warnings across requests in persistent environments via the warningsToBrowser buffer. Warnings are held in a package-level buffer until warningsToBrowser writes them into the page as HTML comments, and the buffer is never cleared between requests. Under mod_perl, FastCGI or PSGI, warnings from one request are written into a later one. Warnings that contain one user's request data can appear in the HTML sent to another user.

CGI versions before 5.00 for Perl allow form fields to be smuggled past proxies and WAFs via multipart parsing that differs from the RFCs

CGI versions before 5.00 for Perl allow form fields to be smuggled past proxies and WAFs via multipart parsing that differs from the RFCs. The multipart/form-data parser matches the boundary anywhere rather than at the start of a line, uses the boundary without its leading -- for some client-supplied User-Agent values, finds the Content-Disposition name and filename inside the quoted values of other parameters, and corrupts parts when the boundary nears the 4096-byte read size.

CGI versions before 5.00 for Perl allow HTTP header injection via entity decoding after the newline check in header and unchecked values in multipart_start and do_push

CGI versions before 5.00 for Perl allow HTTP header injection via entity decoding after the newline check in header and unchecked values in multipart_start and do_push. With autoEscape off, header decodes HTML entities in extra header values after its newline check, so an encoded CR or LF reaches the response. multipart_start does not check header values for newlines, and do_push does not check the type it prints in each part. An application that passes request data into these header values can be made to add headers or split the response.

CGI versions before 5.00 for Perl return upload filenames with any client-supplied path, including ../ and Windows paths

CGI versions before 5.00 for Perl return upload filenames with any client-supplied path, including ../ and Windows paths. The name returned for an uploaded file is the filename parameter of the part's Content-Disposition header, unchanged. An application that saves the upload under this name can be made to write outside the intended directory.

CGI versions before 5.00 for Perl return request-derived values untainted under taint mode

CGI versions before 5.00 for Perl return request-derived values untainted under taint mode. The upload filename, uploadInfo values, some path_info and script_name values, Accept results, and the name lists from url_param and cookie are returned without the taint flag, as is a cookie named expires from 4.67 on. These values come from regex captures or hash keys, which Perl does not taint, so they pass -T checks unvalidated.

CGI::Cookie versions before 5.00 for Perl allow cookie attribute injection via unvalidated path, domain and expires values

CGI::Cookie versions before 5.00 for Perl allow cookie attribute injection via unvalidated path, domain and expires values. The path, domain and expires values are written into the Set-Cookie header without checking for semicolons or control characters. An application that sets them from request data can be made to add other cookie attributes, such as Domain or SameSite.

FFmpeg through 9.0.2 Memory Leak via libx265 Dolby Vision RPU Payloads

FFmpeg through 9.0.2 contains a memory leak vulnerability in libavcodec/libx265.c because libx265_encode_frame() never frees the rpu.payload allocated by ff_dovi_rpu_generate(). Attackers can supply crafted Dolby Vision HEVC or AV1 files that leak one RPU allocation per encoded frame during libx265 transcoding, causing unbounded memory growth and denial of service.

FFmpeg through 9.0.2 NULL Pointer Dereference in G.726 ADPCM Decoder

FFmpeg through 9.0.2 contains a NULL pointer dereference vulnerability in g726_decode_frame() in libavcodec/g726.c because it ignores init_get_bits() failure on oversized packets. Attackers can supply a crafted AVI file with a 268435392-byte G.726 audio chunk to make the decoder read from a NULL buffer, crashing the process.

FFmpeg through 9.0.2 VBLE Decoder NULL Pointer Dereference via vble_decode_frame()

FFmpeg through 9.0.2 contains a NULL pointer dereference vulnerability in vble_decode_frame() in libavcodec/vble.c because it ignores init_get_bits() failure on oversized packets. Attackers can supply a crafted AVI file with a 268435452-byte VBLE frame, causing reads from a NULL bit reader buffer that crash the process.

FFmpeg through 9.0.2 NULL Pointer Dereference in Bink Video Decoder

FFmpeg through 9.0.2 contains a NULL pointer dereference vulnerability in the Bink video decoder because decode_frame() in libavcodec/bink.c ignores init_get_bits() failure. Attackers can supply a crafted Bink file with a frame of about 2^28 bytes, overflowing the bit count and crashing the process for denial of service.