Executive brief
A vulnerability exists in the Spring AI framework's Redis integration, which is used to store and retrieve data for artificial intelligence applications. An attacker could potentially bypass intended data filters to access sensitive information by providing specially crafted search terms. This could lead to unauthorized data exposure from the underlying Redis database.
Technical details
An injection vulnerability exists in the 'spring-ai-redis-store' component of Spring AI. The 'RedisFilterExpressionConverter.stringValue()' method fails to properly escape user-controlled strings before inserting them into the '@field:{VALUE}' RediSearch TAG block. A remote attacker can exploit this by providing malicious filter values, potentially leading to unauthorized data retrieval or manipulation of search logic. The issue is fixed in Spring AI versions 1.0.5 and 1.1.4.
Affected products
- VMware Spring AI 1.0.0 to 1.0.4, 1.1.0 to 1.1.3
Timeline
- 2026-03-27: disclosed: Initial advisory published by VMware
- 2026-03-27: advisory