Data Exposure Flaws Threaten Dify AI Platform Used by 1 Million Apps

Zafran Security has identified four vulnerabilities in the open-source AI platform Dify, which is used in more than one million applications. The flaws could allow attackers to access other tenants' data in multi-tenant cloud configurations.

According to Zafran's advisory, the vulnerabilities stem from improper tenant isolation, potentially enabling an attacker with access to one tenant to read data from another. The issues affect cloud-hosted instances of Dify where multiple tenants share the same infrastructure. Zafran has responsibly disclosed the vulnerabilities to Dify's maintainers.

The disclosure underscores security challenges in open-source AI platforms as adoption scales. Multi-tenant environments require robust isolation mechanisms, and the discovery may prompt Dify to improve its architecture. Dify's maintainers have acknowledged the report and are working on fixes.