Improper sanitization in Microdot's set_cookie() allows header injection.
Microdot is a minimalistic Python web framework. Prior to 2.6.1, the Response.set_cookie() method does not sanitize its string arguments, and in particular will not detect the presence of the \r\n sequence in them. This can be a potential source of header injection attacks. For a header injection attack through this issue to be possible, an attacker must first infiltrate the client (for example through an independent XSS attack), so that it can send malicious information that is destined to be stored in a cookie by the server on behalf of the victim. An attacker that infiltrates one client can only orchestrate a header injection attack for that client, all other clients that were not infiltrated are safe. This vulnerability is fixed in 2.6.1.
pyLoad WebUI leaks Python stack traces to unauthenticated users.
pyLoad is a free and open-source download manager written in Python. Prior to 0.5.0b3.dev100, pyload-ng WebUI returns full Python traceback details to clients on unhandled exceptions. Because /web/<path:filename> is reachable without authentication and renders attacker-controlled template names, an unauthenticated user can reliably trigger a server exception (for example by requesting a non-existent template) and receive internal stack traces in the HTTP response. This vulnerability is fixed in 0.5.0b3.dev100.
pyLoad path traversal allows users with modify perms to write files anywhere.
pyLoad is a free and open-source download manager written in Python. Prior to 0.5.0b3.dev100, when passing a folder name in the set_package_data() API function call inside the data object with key "_folder", there is no sanitization at all, allowing a user with Perms.MODIFY to specify arbitrary directories as download locations for a package. This vulnerability is fixed in 0.5.0b3.dev100.
pyLoad path traversal via insufficient sanitization of folder names.
pyLoad is a free and open-source download manager written in Python. Prior to 0.5.0b3.dev100, package folder names are sanitized using insufficient string replacement. The pattern ....// becomes .._ after replacement (partial removal), leaving .. which can be exploited when the path is later resolved by the OS. This vulnerability is fixed in 0.5.0b3.dev100.
pyLoad allows authenticated users to redirect all traffic via a proxy.
pyLoad is a free and open-source download manager written in Python. Prior to 0.5.0b3.dev100, the set_config_value() API method (@permission(Perms.SETTINGS)) in src/pyload/core/api/__init__.py gates security-sensitive options behind a hand-maintained allowlist ADMIN_ONLY_CORE_OPTIONS. The allowlist contains ("proxy", "username") and ("proxy", "password") — which protect the proxy credentials — but it does not include ("proxy", "enabled"), ("proxy", "host"), ("proxy", "port"), or ("proxy", "type"). Any authenticated user with the non-admin SETTINGS permission can enable proxying and point pyload at any host they control. From that point, every outbound download, captcha fetch, update check, and plugin HTTP call is transparently routed through the attacker. This is a direct continuation of the fix family CVE-2026-33509 / CVE-2026-35463 / CVE-2026-35464 / CVE-2026-35586, each of which patched a different missed option in the same allowlist. This vulnerability is fixed in 0.5.0b3.dev100.
pyLoad allows non-admin users to disable SSL/TLS certificate verification.
pyLoad is a free and open-source download manager written in Python. Prior to 0.5.0b3.dev100, the set_config_value() API method (@permission(Perms.SETTINGS)) in src/pyload/core/api/__init__.py gates security-sensitive options behind a hand-maintained allowlist ADMIN_ONLY_CORE_OPTIONS. The option ("general", "ssl_verify") is not on that allowlist. Any authenticated user with the non-admin SETTINGS permission can set general.ssl_verify = off, and every subsequent outbound pycurl request is made with SSL_VERIFYPEER=0 and SSL_VERIFYHOST=0 — TLS peer and hostname verification are fully disabled. An on-path attacker can then present forged certificates for any hostname pyload fetches. This is a direct continuation of the fix family CVE-2026-33509 / CVE-2026-35463 / CVE-2026-35464 / CVE-2026-35586, each of which patched a different missed option in the same allowlist. This vulnerability is fixed in 0.5.0b3.dev100.
Insecure 'eval' resolver in Hydra config allows arbitrary code execution.
The flash-attention project thru commit e724e2588cbe754beb97cf7c011b5e7e34119e62 (2025-13-04) contains a code injection vulnerability (CWE-94) in its training script. The script registers the Python eval() function as a Hydra configuration resolver under the name eval. This allows configuration files to execute arbitrary Python code via the ${eval:...} syntax. An attacker can exploit this by providing a malicious configuration file, leading to arbitrary code execution when the training script is run with that configuration.
Insecure deserialization in flash-attention checkpoint loading allows RCE.
The flash-attention training framework thru commit e724e2588cbe754beb97cf7c011b5e7e34119e62 (2025-13-04) contains an insecure deserialization vulnerability (CWE-502) in its checkpoint loading mechanism. The load_checkpoint() function in checkpoint.py and the checkpoint loading code in eval.py use torch.load() without enabling the security-restrictive weights_only=True parameter. This allows the deserialization of arbitrary Python objects via the pickle module. An attacker can exploit this by providing a maliciously crafted checkpoint file. When a victim loads this checkpoint during model warmstarting or evaluation, arbitrary code is executed on the victim's system.
CosyVoice insecure deserialization of model files leads to code execution.
CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e (2025-30-21) contains an insecure deserialization vulnerability (CWE-502) in its model loading component. The framework uses torch.load() to load model weight files (e.g., llm.pt, flow.pt, hift.pt) without enabling the security-restrictive weights_only=True parameter. This allows the deserialization of arbitrary Python objects via the pickle module. An attacker can exploit this by providing a malicious model directory containing specially crafted model files. When a victim starts the CosyVoice Web UI pointing to this directory, arbitrary code is executed on the victim's system during the model loading process.
CosyVoice allows RCE via insecure deserialization of malicious torch models.
CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e (2025-30-21) contains an insecure deserialization vulnerability (CWE-502) in its gRPC server component. When the server starts, it loads the speech synthesis model from a user-specified directory using torch.load() without enabling the weights_only=True security parameter. This allows the deserialization of arbitrary Python objects via the pickle module. An attacker can exploit this by providing malicious model files within a directory. When a victim starts the gRPC server pointing to this directory, arbitrary code is executed on the victim's system during server initialization.
Introducing the "VAITP dataset": a specialized repository of Python vulnerabilities and patches, meticulously compiled for the use of the security research community. As Python's prominence grows, understanding and addressing potential security vulnerabilities become crucial. Crafted by and for the cybersecurity community, this dataset offers a valuable resource for researchers, analysts, and developers to analyze and mitigate the security risks associated with Python. Through the comprehensive exploration of vulnerabilities and corresponding patches, the VAITP dataset fosters a safer and more resilient Python ecosystem, encouraging collaborative advancements in programming security.
The supreme art of war is to subdue the enemy without fighting.
Sun Tzu – “The Art of War”
:: Shaping the future through research and ingenuity ::