AI Security Threat Feed

vLLM: RCE via eval() in Qwen3 Coder tool parser

Merlin Transformers4Rec: code injection via Python dep

picklescan: detection bypass allows malicious pickle exec

Keras: safe mode bypass enables RCE via model load

skops: joblib fallback enables RCE via model load

WP Contest Gallery: Stored XSS exposes OpenAI API creds

skops: RCE via MethodNode unsafe deserialization

skops: OperatorFuncNode type confusion → RCE

ExecuTorch: heap overflow in method load, RCE risk

llama_index: path traversal allows arbitrary file read

lollms: timing attack enables credential enumeration

LlamaIndex Obsidian: symlink traversal exposes host files

llama-index Papers Loader: XML expansion DoS

Transformers: ReDoS in chat.py causes CPU exhaustion

Langchain-Chatchat: path traversal exposes system files

Hive Support WP: OpenAI key theft + prompt hijack

jupyter_core: config hijack enables cross-user code exec

Gradio: unauthenticated file copy enables disk DoS

vLLM: image hash collision enables multimodal cache leakage

llama-index-cli: OS command injection enables RCE