Red Hat has announced a Technology Preview container image for the satellite/foreman-mcp-server-rhel9, designed for advanced reporting and AI-driven data analysis within Red Hat Satellite. Among multiple CVEs associated with this advisory, CVE-2026-102267 is detailed as a significant vulnerability in the PyJWT library used for JSON Web Key Set (JWKS) retrieval. This flaw allows an attacker who can manipulate a trusted JWKS endpoint's HTTP redirect responses to redirect key requests to untrusted servers, potentially leading to acceptance of forged cryptographic keys or exposure of sensitive credentials. No patches or mitigations meeting Red Hat's criteria are currently available for this issue. The advisory does not specify affected product versions or confirm active exploitation in the wild.
An unauthenticated remote denial of service vulnerability exists in the Twisted HTTP/2 server due to improper handling of rapid stream resets. A remote client can send a request with END_STREAM followed immediately by a stream reset, causing the server to render the request fully before processing the reset. This allows repeated rapid resets on a single connection to exhaust CPU resources, bypassing the concurrent-stream limit. The issue affects Twisted versions prior to 26.10.0rc1 and has no impact on confidentiality or integrity.
CVE-2026-119120 is a high-severity vulnerability in unclecode's crawl4ai versions 0.9.0 through before 0.9.5. The flaw arises from improper filtering of the CosineStrategy model_name field in Docker API request bodies, allowing authenticated users to specify which HuggingFace model the server loads. This can lead to deserialization of untrusted data, enabling execution of arbitrary Python code under PyTorch versions below 2.6 or resource exhaustion via disk and memory consumption. The issue is fixed in version 0.9.5.
IBM Langflow OSS versions 1.0.0 through 1.12.4 contain a code injection vulnerability (CWE-94) that allows a remote attacker to execute stored Python component code within the server process under the flow owner's runtime context. This bypasses the intended trust boundary restricting public execution paths to trusted server components.
The seapy-ops package on PyPI, versions 1.0.0 through 1.2.2, contains malicious functionality hidden behind undocumented commands. It masquerades as a local file analysis tool but includes backdoor features that exfiltrate files from the user's working directory to an attacker-controlled server and allow remote code execution. The package uploads files recursively, including via symlinks, and can write arbitrary files into the working directory based on server responses. Later versions implement a Python REPL that sends input to the server and executes server-supplied code in memory, disguising errors to conceal activity. These behaviors were confirmed through isolated testing with synthetic credentials and mocked servers, without contacting live endpoints.
The Model Context Protocol (MCP) Python SDK versions 1.10.0 through 1.29.1 and 2.0.0 through 2.1.1 have a vulnerability in server-side bearer-token authentication. The SDK does not verify that a token's audience matches the intended server, allowing tokens issued for other services by the same authorization server to be accepted. This can lead to unauthorized access if tokens meant for different services are accepted. Fixed versions 1.30.0+ and 2.2.0+ introduce a new setting to validate the token audience, but it must be explicitly enabled. Without enabling this validation or manually checking the token audience, the vulnerability remains exploitable.
A vulnerability in python-urllib3 versions prior to 1.26.18-11.oe2403sp3 causes target-server TLS settings to be incorrectly applied to HTTPS proxy connections. This allows an attacker who can intercept and impersonate the HTTPS proxy to observe or modify forwarded traffic or receive a target TLS client certificate. The issue is fixed in version 1.26.18-11.oe2403sp3.
Volmarg Personal Management System versions up to 2.1.0 contain an OS command injection vulnerability. Authenticated users can exploit this flaw by creating directories with shell metacharacters in the upload/files path and then moving or copying them, resulting in arbitrary shell command execution as the web server user.
CVE-2026-118586 is a server-side request forgery (SSRF) vulnerability in the dfm feedfinder2 project affecting versions 0.0.1 through 0.0.4. The flaw exists in the find_feeds function of feedfinder2.py and allows remote attackers to manipulate requests, potentially causing the server to make unintended HTTP requests. The vulnerability has a medium severity rating with a CVSS score of 6.9. The project maintainers have been informed but have not yet responded or released a fix.
Tornado HTTP servers prior to version 6.6 are vulnerable to event loop blocking and excessive memory use when processing HTTP/1.1 chunked transfer encoded bodies composed of many small chunks. An attacker can send a request body with millions of 1-byte chunks, causing the server's event loop to block for extended periods (up to a minute for bodies near 100 MB), impacting all connections served by the process. This vulnerability also affects Tornado's SimpleAsyncHTTPClient when reading responses from malicious servers. The issue is fixed in Tornado 6.6, which introduces periodic yielding to the event loop and more efficient chunk handling.
Electrum versions prior to 4.8.3 are vulnerable to a malicious server attack where the server can replace a just-broadcast transaction with a conflicting transaction containing the attacker's output address. This attack exploits the client's trust in the server's mempool data and the lack of signature verification for unconfirmed transactions, potentially tricking users into fee-bumping a fraudulent transaction and losing funds. The issue is fixed in Electrum 4.8.3 by verifying signatures of wallet-owned inputs in mempool transactions before adding them to the local history.
vLLM versions prior to 0.32.0 have a permissive default CORS configuration allowing any origin to send cross-origin requests to the HTTP server. Without an API key configured, no authentication is enforced, enabling malicious web pages to access and interact with the server if reachable by the victim's browser. This can lead to unauthorized enumeration of models and execution of chat completions, potentially consuming GPU resources and exposing model outputs. The issue is fixed in version 0.32.0.
The Phi-4 multimodal audio encoder in vLLM versions prior to 0.32.0 runs its conformer attention in eager mode by default, which fully materializes and retains large attention matrices per layer during audio processing. This causes GPU memory to grow monotonically within a single request, leading to out-of-memory errors and a crash of the entire server when processing a 10-minute audio clip. The retained attention matrices persist beyond request completion, causing sustained high memory usage. This vulnerability allows any unauthenticated client with API access to cause a full denial of service with a single legal audio request.
vLLM versions 0.19.0 through 0.31.x have a denial-of-service vulnerability in their chat template rendering. The issue arises because the Jinja2 rendering in a sandboxed environment lacks execution timeouts and output limits, and runs in a single-threaded pool that cannot be interrupted. An authenticated attacker can submit a crafted template that exhausts the renderer thread pool or causes out-of-memory conditions, permanently denying chat-style API endpoints until the process is restarted. This affects all tenants on the same API-server replica and can also be triggered by malicious model repositories with hostile templates. The vulnerability is fixed in version 0.32.0.