Four Vulnerabilities in Dify Allow Unauthenticated Read of AI Conversions

Cybersecurity researchers have disclosed details of four vulnerabilities in Dify, an open‑source agentic workflow platform with more than 146,000 GitHub stars, that could allow attackers to stealthily read artificial intelligence (AI) conversions from other customers' applications without requiring authentication.

The vulnerabilities, collectively codenamed DifyTap by Zafran Security, enable unauthenticated users to access sensitive AI conversion data, potentially exposing proprietary models and user inputs. DifyTap exploits a combination of improper access controls and insecure data handling across multiple components of the platform. The discovery was made after a thorough audit of the codebase and the exploitation was demonstrated in a proof‑of‑concept environment. Dify has issued a patch that addresses all four issues by tightening authentication checks and sanitizing data flows. Users of the platform should upgrade to the latest release immediately to prevent data leakage. The incident highlights the importance of rigorous access control in AI‑driven services.