1 CRIT · 6 HIGH · 2 MED · 3 INFO · THREAT RED · 12 items · Generated in 231s
A patch bypass allows unauthenticated remote code execution on default Flowise deployments via MCP package auto-install.
Unsanitized input lets attackers inject Stata commands and reach OS command execution on the host account.
Scoped reads can leak one tenant's stored data into another tenant's results due to namespace prefix collisions.
Authenticated users can bypass URL filters to access internal services or cloud metadata and exfiltrate responses via the API.
Unauthenticated regex input can trigger CPU exhaustion and stall structured-output requests on exposed vLLM servers.
A crafted username can run arbitrary OS commands on the MCP host with the server user's privileges.
DNS rebinding can bypass SSRF checks and let attackers reach internal or metadata services through the test endpoint.
Unauthenticated path traversal can expose session log JSONL files and leak sensitive data from the host filesystem.
Shows vendor focus on controlling AI agent access through unified data and identity governance.
Default autonomous mode expands agent use, but this item is product news rather than a confirmed vulnerability or exploit.
This corrects attribution of a large compromise and prevents wasting response effort on the wrong package or root cause.
Stolen OAuth tokens can bypass phishing defenses and grant access to Gmail, Drive, and linked systems.