A security vulnerability in Python's http.cookiejar.CookieJar module allowed a server to delete cookies belonging to other domains from a shared cookie jar by sending an expired Set-Cookie header with manipulated Domain and Path attributes. This behavior violated the cookie policy as specified in RFC 6265 section 5.3. The issue was fixed in Python version 3.9.9-64.oe2203sp4 by ensuring the cookie policy is consulted before deleting existing cookies.
CVE-2026-73971 is a code injection vulnerability in Music Assistant server versions prior to 2.8.0. The flaw arises from unsafe use of Python's eval() on user-controlled input in the parse_arguments() function, allowing authenticated users to execute arbitrary Python code with server process privileges. This can lead to exposure of secrets, alteration of media library data, and full control over the application container. The issue is fixed in version 2.8.0.
The solana-py-utils package on PyPI contains malicious code that executes during installation. It runs a bundled script that steals sensitive data including AWS credentials, SSH keys, various configuration files, environment variables with secrets, and cryptocurrency wallet data from browser extensions. The stolen data is exfiltrated to a hardcoded remote server. The package name is a typosquat of legitimate Solana tooling, aiming to trick users into installing it.
The anchorpy-utils package on PyPI is a malicious typosquatting package that executes harmful code during installation. It exfiltrates sensitive data including SSH keys, AWS credentials, npm configuration, Solana wallet files, environment variables containing secrets, and browser extension data related to cryptocurrency wallets. The stolen data is sent to a hardcoded external server. The package falsely claims to be related to the Solana Python client but provides no legitimate functionality.
SharpIppNext is a .NET Standard library for building Internet Printing Protocol clients and servers. Prior to 4.1.0, server-side SharpIppNext implementations using SharpIppServer.ReceiveRequestAsync or SharpIppServer.ReceiveRawRequestAsync pass incoming data to IppProtocol.ReadIppRequestAsync in SharpIpp/Protocol/IppProtocol.cs, where ReadDocumentStream is enabled by default and the remaining document body is copied with CopyToAsync into a MemoryStream without a size limit. An unauthenticated remote client can send an arbitrarily large IPP document body, exhausting process memory and causing an out-of-memory failure or service termination. Client-only implementations using SharpIppClient are not affected. This issue is fixed in version 4.1.0.
Cisco Talos identified a financially motivated threat group, UAT-12643, using multiple payloads including the Go-based Overlord RAT, a Python infostealer called Good Game, and a cryptocurrency clipboard hijacker. The threat actor targets cryptocurrency holders, gaming platform users, VoIP software users, and finance/accounting staff, primarily in Europe. The Overlord RAT uses a novel command-and-control resilience technique by embedding encrypted server URLs in Solana blockchain transaction memos. Initial access is gained via weaponized Windows shortcuts, HTA files, and MSBuild XML files. The actor exfiltrates stolen data using Discord webhooks and uploads to gofile.io. The campaign shows low-confidence links to North Korea-aligned APT actors based on TTP overlaps and use of Overlord RAT variants previously seen in North Korean campaigns.
Malicious versions 1.26.1 and 1.26.2 of the legitimate Python package agent-framework (Microsoft Agent Framework) were published on PyPI containing a credential-stealing backdoor. These releases add a .pth startup hook that executes code on every Python interpreter start, harvesting environment variables, credential files, cloud tokens, and other sensitive data. The stolen data is compressed, encoded, and exfiltrated via HTTP POST to a remote server. The malicious versions have been removed from PyPI. Versions 1.26.0 and earlier are not affected. Users who installed the compromised versions should consider their environments exposed and rotate all accessible credentials and secrets.
pympp version 0.8.0 contains a vulnerability where the fee_payer server validates calldata leniently, allowing attackers to append padding bytes that increase the intrinsic gas cost without violating gas limit or fee caps. This leads to excessive sponsored transaction fees, inflating operational costs significantly. The issue is fixed in version 0.9.1.
A critical format string injection vulnerability exists in the applyChatTemplate GraphQL endpoint of arize-phoenix versions 13.0.0 through before 15.7.0. This flaw allows attackers to read sensitive server-side environment variables, including cryptographic secrets and internal module attributes, by exploiting improper attribute access control in template rendering. On default installations without authentication, the endpoint is accessible without credentials; on authenticated deployments, any user account can exploit it. Successful exploitation enables extraction of the PHOENIX_SECRET master key, allowing forging of admin JWT tokens and decryption of stored secrets.
ESP-IDF is the Espressif Internet of Things (IoT) Development Framework. Prior to 5.2.8, 5.3.6, 5.5.6, and 6.0.3, the Bluedroid BLE Mesh GATT server read handler bt_mesh_bta_gatts_cb in esp_ble_mesh stores a signed attribute-read callback length in an unsigned 16-bit variable. A negative ATT error code is therefore interpreted as a very large successful length and used for a memory copy from a small stack buffer into the GATT response structure. A remote attacker within Bluetooth LE range who can connect to an exposed PB-GATT provisioning or Mesh Proxy service can trigger a failed attribute read with a crafted request. On devices where the affected attributes are accessible without pairing, exploitation requires neither pairing nor user interaction; access requirements depend on device configuration. The resulting out-of-bounds read and stack-based buffer overflow corrupts the Bluetooth host task stack, causing a crash and device restart and potentially further memory corruption. The issue affects devices built with CONFIG_BT_BLUEDROID_ENABLED=y and BLE Mesh GATT enabled; the NimBLE host path is not affected. This issue is fixed in versions 5.2.8, 5.3.6, 5.5.6, and 6.0.3.
The mcp-aevo-server package on PyPI is a malicious package that targets users of ribbon-finance/aevo-trading-skills by exfiltrating environment variable credentials and executing remote commands via a listening thread. It employs dependency confusion techniques and uses a Telegram bot for command and control.
Moodle versions up to 5.3.0, including specific subversions of 5.2, 5.1, 5.0, and 4.5, contain a path traversal vulnerability in the Blackboard question import feature. This flaw allows teachers with import or lesson edit permissions to upload crafted zip files that exploit directory traversal sequences to copy sensitive server files such as config.php into their draft area.
A vulnerability in AIOHTTP's WebSocket implementation allows attackers to bypass receive queue limits by sending many small or compressed messages. This can cause excessive memory consumption on the server or client side. The issue is fixed in version 2.19.0_2.
A vulnerability in aiohttp's multipart handling allows bypassing the client_max_size limit when processing parts with Content-Transfer-Encoding set to base64. This causes the server to read excessive data into memory or spin the event loop, potentially leading to denial of service. The issue affects versions >=2.19.0 and fixed in 2.19.0_2.
AIOHTTP's client CookieJar component improperly parses Set-Cookie headers, creating additional cookies from unrecognized name=value attributes without limiting the number of stored cookies. This allows a malicious server to cause excessive memory and CPU consumption on the client by forcing it to store and send very large Cookie headers, especially on repeated redirects. The issue affects scrapy versions from 2.19.0 up to but not including 2.19.0_2. A patch is available in version 2.19.0_2.
AIOHTTP's Application.add_domain() function had a performance issue where using two or more wildcards (*) in a domain mask caused polynomial-time matching delays on long Host headers. This could allow an attacker to stall the server's event loop by sending crafted requests with long Host headers against such domain masks. The issue is fixed in version 2.19.0_2. Until patched, a workaround is to limit domain masks to at most one wildcard.
Since July 2025, a sophisticated cyberespionage campaign has targeted Southeast Asian government entities using two custom remote access Trojans (PerfectPioneer RAT and KUKURAT). These RATs use covert HTTP/3 over UDP for command and control, bypassing traditional TCP detection. The campaign employs advanced stealth techniques including operational safety mechanisms and heuristic camouflage. Initial access is gained via server exploitation and phishing, followed by credential harvesting and lateral movement. Sensitive documents and emails are exfiltrated with high operational stealth and ongoing capability refinement.
A vulnerability in Jupyter Server version 4.0 causes 5xx error request logs to leak Referer header values that contain sensitive tokens. This could expose token-bearing URLs in server logs.