CVE Vulnerability Catalog

savg-sanitizer is a PHP SVG/XML sanitizer. Prior to 1.0.0, isHrefSafeValue() validates an SVG href after XML DTD entity expansion, but saveXML() serializes the original entity reference after removing the DTD declaration. A crafted entity such as Tab can appear to the sanitizer as a safe fragment prefix while HTML5 Named Character Reference resolution during inline HTML rendering later converts the surviving reference to whitespace, exposing a javascript: URL. When an application embeds the sanitized SVG inline, a user who activates the link can cause script to execute in the embedding page's origin. This issue is fixed in version 1.0.0.

savg-sanitizer is a PHP SVG/XML sanitizer. Prior to 1.0.0, a crafted SVG DTD with a #FIXED attribute default can make cleanAttributesOnWhitelist() perform a double DOMElement::removeAttribute() call on the same attribute name in src/Sanitizer.php. The first removal deletes the explicit attribute, while the DTD default rematerializes the value before the href safety path performs the second removal, which can corrupt libxml state and terminate the PHP worker. An attacker who can submit SVG content to a sanitization endpoint can repeatedly interrupt workers and degrade or exhaust application availability. This issue is fixed in version 1.0.0.

The Arkime live capture service (arkime-live) in Malcolm runs with network_mode: host, exposing port 8005 on all network interfaces (viewHost=0.0.0.0). Arkime trusts the X-Forwarded-User header from any IP address (userAuthIps=::,0.0.0.0/0) and auto-creates users with full access. The passwordSecret is hardcoded to the public value "Malcolm". A network-adjacent attacker bypasses nginx entirely by connecting directly to port 8005 with a forged identity header.

Malcolm's front nginx reverse proxy defines a "Dashboards → Arkime shortcut" location using a case-insensitive regex matcher but a case-sensitive rewrite. A request whose path segment is not exact-lowercase (for example /IDDASH2ARK/...) enters the location (the matcher fires) but evades the rewrite (no redirect is issued), so nginx falls through to the location's proxy_pass to the Arkime backend. That location is the one proxied location in the shipped config that does not include the per-location authentication file, so the request reaches Arkime unauthenticated. The same location also forwards a client-supplied X-Forwarded-User header un-overwritten, and Arkime is configured to trust X-Forwarded-User as the authenticated username — so an unauthenticated network caller can reach the Arkime backend while supplying a forged, auto-provisioned identity.

Malcolm's upload-processing pipeline (scripts/safe-extract.py) enforces entry-count, nesting-depth, and total-uncompressed-byte limits when extracting container archives (zip/tar/rar/7z via libarchive), but those limits are not applied when the uploaded file is a single-stream compressed format (.gz, .bz2, .xz, .lzma, .lz) that isn't a .tar.*-style archive. Any authenticated user permitted to upload PCAP/log files can upload a small, highly compressible file (e.g. a gzip bomb) that decompresses to an effectively unbounded size on disk, exhausting the shared Docker volume used by OpenSearch, Logstash, Arkime, and Zeek, and disrupting the platform for all users.

Malcolm's nginx Lua role-based access control (RBAC) layer decides whether an authenticated user may reach a role-restricted path (e.g. /htadmin, /auth, /admin_login, /arkime/api/esadmin, NetBox, upload endpoints) by pattern-matching the raw, percent-encoded request URI. Nginx itself, however, selects which location block actually serves the request using the percent-decoded, normalized URI. Because the RBAC check never percent-decodes its input, an authenticated low-privilege user can request an admin-only path using percent-encoding (e.g. /%68tadmin.php) and have nginx route it to the restricted location while the Lua RBAC gate evaluating the un-decoded raw string finds no matching restriction and grants access.

Insecure file permissions in the CodeCatalyst connection handler in AWS Toolkit for VS Code before 4.10.0 allowed local users to obtain CodeCatalyst bearer tokens via reading world-readable token cache files. To mitigate this issue, users should upgrade to version 4.10.0 or later.

GnuTLS v3.8.13 may unexpectedly trust a certificate based on Common Name when Subject Alternative Name contains an SRV otherName (1.3.6.1.5.5.7.8.7) with an OCTET STRING value. The supplier does not consider this an actionable vulnerability due to its threat model.

The 'github.com/moby/sys/user' package provides Go utilities for parsing and looking up entries in Unix-style user and group database files. Versions before 0.4.1 do not sufficiently limit entries when parsing '/etc/passwd'- or '/etc/group'-style files, allowing an attacker who can supply a specially crafted file to cause excessive memory consumption and potentially terminate the affected process due to an out-of-memory condition.

A policy-enforcement flaw in Zimbra Collaboration Suite allows an authenticated user to bypass disabled mail forwarding by using a Sieve notify action to send copies of email content and headers to an arbitrary address.

An authorization bypass in the EWS FindItem handler of Zimbra Collaboration Suite 10.1.0 through 10.1.19 allows an authenticated user with EWS enabled to read complete mailbox items, including raw MIME and attachments, from arbitrary local accounts without a share or delegation grant.

msgpack5 is a msgpack v5 implementation for node.js and the browser. Prior to 6.1.0, constructing msgpack5 with an empty or partial options object disables the default protoAction: 'error' protection. A decoded map containing a __proto__ key can then replace the decoded object's prototype, potentially changing inherited properties or downstream behavior, although Object.prototype is not modified globally.

msgpack5 before 6.1.0 treats the reserved byte 0xc1 as incomplete input instead of invalid. When a stream begins with 0xc1, subsequent data remains buffered while waiting for bytes that cannot make the value valid, allowing a remote peer to exhaust memory.

msgpack5 before 6.1.0 has no nesting-depth limit for arrays and maps, allowing an attacker who can provide MessagePack input to submit deeply nested containers that exhaust the JavaScript call stack and interrupt a process, worker, or request handler.

In msgpack5 before 6.1.0, the streaming decoder reparses an incomplete array or map from the beginning whenever another chunk arrives. A remote peer can split one valid MessagePack container across many small chunks, causing completed elements to be decoded repeatedly, producing quadratic CPU use and blocking the event loop.

Pydantic AI from 1.77.0 until 1.107.2 and from 2.24.0 until 2.24.0? Actually: until 1.107.2 and 2.24.0, web_fetch_tool, the WebFetch local fallback, and remote FileUrl media downloads buffer the complete HTTP response body before enforcing content-size controls. An attacker-influenced URL can stream an arbitrarily large response that exhausts process memory and crashes the worker.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 0.3.4 until 1.107.4 and 2.27.1, OpenTelemetry instrumentation configured with InstrumentationSettings(include_content=False) can export retry prompts outside tool calls in gen_ai.input.messages and pydantic_ai.all_messages. Agents using NativeOutput, PromptedOutput, or output validators on text output can therefore disclose validation feedback, including invalid model values quoted by that feedback, to readers of the telemetry backend. Tool-call retries and deployments that do not use include_content=False are not affected by this specific path. This issue is fixed in versions 1.107.4 and 2.27.1.

Pydantic AI from 1.34.0 until 1.107.5 and from 2.24.0 until 2.30.0, the Agent.to_web() and clai web development chat server does not validate the Host header, allowing a website visited by a developer to use DNS rebinding to reach a loopback-hosted agent as a same-origin service. The hostile page can read the served UI and submit chat requests that execute agent tools with the local process's privileges and credentials.

Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 0.3.4 until 1.107.6 and 2.44.0, OpenTelemetry instrumentation configured with InstrumentationSettings(include_content=False) can still export sensitive agent content through exception.message and exception.stacktrace events, error status descriptions, and model_request_parameters containing instructions or the prompted_output_template. The exposed data is available to readers of the configured telemetry backend and can include tool feedback, provider error bodies, runtime instructions, and structured-output templates even though message attributes are redacted. This issue does not grant new access to agent data, and deployments that do not use include_content=False are not affected by the setting bypass. This issue is fixed in versions 1.107.6 and 2.44.0.

Pydantic AI from 1.77.0 until 1.107.6 and from 2.24.0 until 2.44.0, the local web_fetch_tool and the WebFetch local fallback process server-controlled responses with quadratic title extraction, whitespace normalization, and ordered-list numbering. An attacker-controlled page of modest size can therefore block the event loop for an extended period, stalling other agent runs and requests, while unsupported codecs or excessive HTML or JSON nesting can abort an individual run.