CVEs

There is an information disclosure vulnerability in Esri Portal for ArcGIS versions 12.0 and earlier that may under difficult to reproduce circumstances allow a remote, unauthenticated attacker to reflect sensitive information in a http response body.

Hydra is a framework for elegantly configuring complex applications. Prior to 1.3.4, hydra.utils.instantiate() resolves and calls Python objects selected by configuration through _resolve_target() in hydra/_internal/instantiate/_instantiate2.py, allowing attacker-controlled target values and arguments to choose dangerous callables. A consuming application, library, CLI workflow, or model loader that passes untrusted configuration, CLI overrides, or model metadata into hydra.utils.instantiate() can therefore execute arbitrary code in its own process, including reading or modifying files and credentials or terminating the process. Version 1.3.4 adds target blocking with an explicit HYDRA_INSTANTIATE_ALLOWLIST_OVERRIDE escape hatch. This issue is fixed in version 1.3.4.

Rejected reason: This CVE ID has been rejected or withdrawn by its CVE Numbering Authority.

Atlantis is a self-hosted golang application that listens for Terraform pull request events via webhooks. From 0.19.8 until 0.45.0, Atlantis does not consistently validate user-controlled workspace values supplied through accepted repository-level atlantis.yaml configuration or authenticated /api/plan input before joining them into local workspace paths. Traversal segments can escape the intended per-pull workspace directory and cause clone preparation or other working-directory code paths to call os.RemoveAll, os.MkdirAll, or related filesystem operations on out-of-bounds directories before Terraform rejects the invalid workspace name. This can create, delete, or reuse writable paths with the privileges of the Atlantis process, causing integrity loss or denial of service. This issue is fixed in version 0.45.0.

Keystone is a content management system for Node.js. Prior to 6.5.3, the findMany resolver in packages/core/src/lib/core/queries/resolvers.ts compares the signed take argument directly with graphql.maxTake, allowing a remote unauthenticated GraphQL client to provide a negative take value whose magnitude exceeds the configured bound. The bypass also applies to relationship queries and can return more records than the developer intended, potentially exhausting service resources. This issue is fixed in version 6.5.3.

YOURLS is a self-hosted, customizable URL shortener written in PHP. From 1.5.1 until 1.10.4, YOURLS stores the HTTP Referer header through yourls_get_referrer(), yourls_sanitize_url_safe(), and yourls_log_redirect(), then aggregates the value in yourls-infos.php and passes the derived domain through yourls_get_domain(), yourls_stats_pie(), and yourls_google_array_to_data_table(). The chart builder concatenates labels into inline JavaScript without JavaScript-string escaping, so an unauthenticated attacker can poison the statistics of an existing short URL with a crafted referrer. When an administrator or public stats-page viewer opens the affected statistics page, attacker-controlled JavaScript executes in the YOURLS origin and can access admin-visible data, the API signature token, and privileged same-origin actions. This issue is fixed in version 1.10.4.

Microsoft UFO open-source framework for intelligent automation across devices and platforms. Prior to 3.0.8, ufo/client/mcp/http_servers/linux_mcp_server.py binds a FastMCP streamable HTTP server to localhost:8010 but does not validate the Host, Origin, or Sec-Fetch-Site headers. An attacker-controlled web page can use DNS rebinding to reach the local /mcp endpoint, enumerate tool schemas through tools/list, and invoke execute_command with a valid UFO_MCP_API_KEY to read files or execute allowed operating system commands as the victim's user. This issue is fixed in version 3.0.8.

Nezha Monitoring is a self-hostable, lightweight, servers and websites monitoring and O&M tool. Nezha versions 1.14.13 through 1.14.14 and 2.0.0 through 2.0.9 do not bind stream identifiers created by CreateStream in service/rpc/io_stream.go to their creating user, and `GET /ws/terminal/:id` and `GET /ws/file/:id` only check whether the supplied UUID exists. An authenticated RoleMember who obtains a live stream UUID from logs, browser history, referer data, or telemetry can attach to another user's terminal or file-manager session, read and write target-server files, and execute shell commands. This issue is fixed in version 2.0.10.

Defuddle cleans up HTML pages. Prior to 0.19.1, site extractors interpolate page-derived image alt and src values, og:image values, and video descriptions into HTML strings without context-appropriate escaping, and buildExtractorResponse() returns this contentHtml without the main pipeline's DOM-based sanitization. The affected paths include src/extractors/x-article.ts, src/extractors/substack.ts, and src/extractors/youtube.ts. A malicious page or attacker-controlled content on a matching domain can inject event-handler attributes or javascript URLs that execute when a victim or downstream application renders the extracted HTML. This issue is fixed in version 0.19.1.

Xinference is an inference API for running open-source, speech, and multimodal models. In 2.5.0 and earlier, Xinference passes attacker-influenced Llama3 tool-call output to eval() in xinference/model/llm/tool_parsers/llama3_tool_parser.py and xinference/model/llm/utils.py. Requests to /v1/chat/completions with a tools field flow through xinference/api/restful_api.py, xinference/model/llm/transformers/core.py, handle_chat_result_non_streaming(), and _post_process_completion() before extract_tool_calls() or _eval_llama3_chat_arguments() evaluates the model-generated Python expression. An unauthenticated remote attacker can influence that output through a crafted prompt and execute commands in the Xinference server process context. This issue is fixed in version 2.7.0.