ReDoS in multipart 1.3.0 - `parse_options_header()`
multipart is a fast multipart/form-data parser for python. Prior to 1.2.2, 1.3.1 and 1.4.0-dev, the parse_options_header() function in multipart.py uses a regular expression with an ambiguous alternation, which can cause exponential backtracking (ReDoS) when parsing maliciously crafted HTTP or multipart segment headers. This can be abused for denial of service (DoS) attacks against web applications using this library to parse request headers or multipart/form-data streams. The issue is fixed in 1.2.2, 1.3.1 and 1.4.0-dev.
SGLangs `replay_request_dump.py` contains an insecure pickle.load() without validation and proper deserialization. An attacker can take advantage of this by providing a malicious .pkl file, which will execute the attackers code on the device running the script.
SGLang' encoder parallel disaggregation system is vulnerable to unauthenticated remote code execution through the disaggregation module, which deserializes untrusted data using pickle.loads() without authentication.
SGLang's multimodal generation module is vulnerable to unauthenticated remote code execution through the ZMQ broker, which deserializes untrusted data using pickle.loads() without authentication.
ha-mcp has XSS via Unescaped HTML in OAuth Consent Form
ha-mcp is a Home Assistant MCP Server. Prior to 7.0.0, the ha-mcp OAuth consent form renders user-controlled parameters via Python f-strings with no HTML escaping. An attacker who can reach the OAuth endpoint and convince the server operator to follow a crafted authorization URL could execute JavaScript in the operator's browser. This affects only users running the beta OAuth mode (ha-mcp-oauth), which is not part of the standard setup and requires explicit configuration. This vulnerability is fixed in 7.0.0.
ha-mcp OAuth 2.1 DCR mode enables network reconnaissance via an error oracle
ha-mcp is a Home Assistant MCP Server. Prior to 7.0.0, the ha-mcp OAuth consent form (beta feature) accepts a user-supplied ha_url and makes a server-side HTTP request to {ha_url}/api/config with no URL validation. An unauthenticated attacker can submit arbitrary URLs to perform internal network reconnaissance via an error oracle. Two additional code paths in OAuth tool calls (REST and WebSocket) are affected by the same primitive. The primary deployment method (private URL with pre-configured HOMEASSISTANT_TOKEN) is not affected. This vulnerability is fixed in 7.0.0.
Copyparty has unexpected JavaScript execution via crafted URL to folder with `.prologue.html`
Copyparty is a portable file server. Prior to 1.20.12, if an attacker has been given both read- and write-permissions to the server, they can upload a malicious file with the filename .prologue.html and then craft a link to potentially execute arbitrary JavaScript in the victim's context. Note that it is intended behavior that the JavaScript would execute if the target clicks a link to the HTML file itself; "https://example.com/foo/.prologue.html". The vulnerability is that "https://example.com/foo/?b" would also evaluate the file, making the behavior unexpected. There are existing preventative measures (strict SameSite cookies) which makes i
Copyparty ftp/sftp: Sharing a single file did not fully restrict source-folder access
Copyparty is a portable file server. Prior to 1.20.12, there was a missing permission-check in the shares feature (the shr global-option). This vulnerability only applies when the shares feature is used for the specific purpose of creating a share of just a single file inside a folder or either the FTP or SFTP server is enabled, and also made publicly accessible. Given these conditions, when a user is browsing a share through either FTP or SFTP (not http or https), they can gain read-access to the remaining files inside the shared folder by guessing/bruteforcing the filenames. It was not possible to descend into subdirectories in this manner;
Tornado is a Python web framework and asynchronous networking library. In versions of Tornado prior to 6.5.5, the only limit on the number of parts in multipart/form-data is the max_body_size setting (default 100MB). Since parsing occurs synchronously on the main thread, this creates the possibility of denial-of-service due to the cost of parsing very large multipart bodies with many parts. This vulnerability is fixed in 6.5.5.
Black's vulnerable version parsing leads to RCE in GitHub Action
Black is the uncompromising Python code formatter. Black provides a GitHub action for formatting code. This action supports an option, use_pyproject: true, for reading the version of Black to use from the repository pyproject.toml. A malicious pull request could edit pyproject.toml to use a direct URL reference to a malicious repository. This could lead to arbitrary code execution in the context of the GitHub Action. Attackers could then gain access to secrets or permissions available in the context of the action. Version 26.3.0 fixes this vulnerability.
pypdf: manipulated stream length values can exhaust RAM
pypdf is a free and open-source pure-python PDF library. Prior to 6.8.0, an attacker who uses this vulnerability can craft a PDF which leads to large memory usage. This requires parsing a content stream with a rather large /Length value, regardless of the actual data length inside the stream. This vulnerability is fixed in 6.8.0.
django-unicorn affected by component state manipulation via unvalidated attribute access
Unicorn adds modern reactive component functionality to your Django templates. Prior to 0.67.0, component state manipulation is possible in django-unicorn due to missing access control checks during property updates and method calls. An attacker can bypass the intended _is_public protection to modify internal attributes such as template_name or trigger protected methods. This vulnerability is fixed in 0.67.0.
MCP Atlassian has an arbitrary file write leading to arbitrary code execution via unconstrained download_path in confluence_download_attachment
MCP Atlassian is a Model Context Protocol (MCP) server for Atlassian products (Confluence and Jira). Prior to version 0.17.0, the `confluence_download_attachment` MCP tool accepts a `download_path` parameter that is written to without any directory boundary enforcement. An attacker who can call this tool and supply or access a Confluence attachment with malicious content can write arbitrary content to any path the server process has write access to. Because the attacker controls both the write destination and the written content (via an uploaded Confluence attachment), this constitutes for arbitrary code execution (for example, writing a vali
Высокий · CVSS 8.2EPSS 13%Публичный эксплойтЕсть исправление
MCP Atlassian has SSRF via unvalidated X-Atlassian-Jira-Url / X-Atlassian-Confluence-Url headers
MCP Atlassian is a Model Context Protocol (MCP) server for Atlassian products (Confluence and Jira). Prior to version 0.17.0, an unauthenticated attacker who can reach the mcp-atlassian HTTP endpoint can force the server process to make outbound HTTP requests to an arbitrary attacker-controlled URL by supplying two custom HTTP headers without an `Authorization` header. No authentication is required. The vulnerability exists in the HTTP middleware and dependency injection layer — not in any MCP tool handler - making it invisible to tool-level code analysis. In cloud deployments, this could enable theft of IAM role credentials via the instance
Copyparty volflag `nohtml` did not block javascript in svg files
Copyparty is a portable file server. Prior to v1.20.11., the nohtml config option, intended to prevent execution of JavaScript in user-uploaded HTML files, did not apply to SVG images. A user with write-permission could upload an SVG containing embedded JavaScript, which would execute in the context of whichever user opens it. This has been fixed in v1.20.11.
Glances has SQL Injection via Process Names in TimescaleDB Export
Glances is an open-source system cross-platform monitoring tool. Prior to 4.5.1, The TimescaleDB export module constructs SQL queries using string concatenation with unsanitized system monitoring data. The normalize() method wraps string values in single quotes but does not escape embedded single quotes, making SQL injection trivial via attacker-controlled data such as process names, filesystem mount points, network interface names, or container names. This vulnerability is fixed in 4.5.1.
Glances is an open-source system cross-platform monitoring tool. Prior to 4.5.1, the /api/4/config REST API endpoint returns the entire parsed Glances configuration file (glances.conf) via self.config.as_dict() with no filtering of sensitive values. The configuration file contains credentials for all configured backend services including database passwords, API tokens, JWT signing keys, and SSL key passwords. This vulnerability is fixed in 4.5.1.
vLLM is an inference and serving engine for large language models (LLMs). The SSRF protection fix for CVE-2026-24779 add in 0.15.1 can be bypassed in the load_from_url_async method due to inconsistent URL parsing behavior between the validation layer and the actual HTTP client. The SSRF fix uses urllib3.util.parse_url() to validate and extract the hostname from user-provided URLs. However, load_from_url_async uses aiohttp for making the actual HTTP requests, and aiohttp internally uses the yarl library for URL parsing. This vulnerability in 0.17.0.
Arbitrary File Read via Absolute Path Input in nltk.util.filestring()
A vulnerability in the `filestring()` function of the `nltk.util` module in nltk version 3.9.2 allows arbitrary file read due to improper validation of input paths. The function directly opens files specified by user input without sanitization, enabling attackers to access sensitive system files by providing absolute paths or traversal paths. This vulnerability can be exploited locally or remotely, particularly in scenarios where the function is used in web APIs or other interfaces that accept user-supplied input.