CVE-2026-108664 in JeecgBoot
Summary
by MITRE • 10/11/2026
JeecgBoot through 3.9.5 contains a missing authorization vulnerability in the AiragPromptsController queryById handler that allows low-privileged authenticated users to read any AI prompt template. Attackers can enumerate ids via the unguarded /airag/prompts/list endpoint and query each one to obtain prompt text, model parameters, and creator details belonging to administrators or other users.
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 10/11/2026
The vulnerability identified in JeecgBoot versions up to 3.9.5 represents a critical failure in access control mechanisms within the AiragPromptsController module. Specifically, the queryById handler lacks proper authorization checks, allowing any authenticated user with low privileges to retrieve sensitive data associated with AI prompt templates. This flaw enables unauthorized actors to bypass intended security boundaries and access information that should be restricted to administrators or specific authorized personnel. The core technical issue lies in the absence of role-based verification before processing requests for individual resource instances, which violates fundamental principles of secure software design regarding object-level permissions.
Attackers can exploit this vulnerability by first enumerating valid identifiers through the unguarded /airag/prompts/list endpoint. This listing function does not enforce access controls on its output, allowing low-privileged users to discover a comprehensive list of prompt IDs and associated metadata such as creator details. Once these identifiers are obtained, the attacker utilizes them in subsequent requests directed at the queryById handler. Because this specific handler fails to validate whether the requesting user has permission to view the target resource, it returns full details including the prompt text, model parameters, and other configuration data. This two-step exploitation path effectively transforms a simple enumeration flaw into a significant information disclosure vulnerability.
The operational impact of this vulnerability is substantial, particularly in environments where AI prompts contain sensitive business logic, proprietary algorithms, or internal system configurations. The exposure of creator details can facilitate further social engineering attacks by revealing the identities and potentially contact information of administrators or key developers. Furthermore, access to model parameters may allow attackers to understand the underlying architecture of the AI integration, aiding in more sophisticated adversarial machine learning attacks such as prompt injection or data poisoning attempts. In a multi-tenant SaaS context, this could lead to cross-tenant data leakage if tenant isolation is not strictly enforced at other layers of the application stack.
This vulnerability aligns with CWE-284, which describes Improper Access Control, specifically highlighting failures in enforcing restrictions on authenticated users' access to resources. It also maps closely to MITRE ATT&CK technique T1078, Valid Accounts, as it relies on legitimate but insufficiently privileged credentials to perform unauthorized data retrieval. Additionally, the enumeration aspect relates to CWE-209, Information Exposure Through an Error Message or Behavioral Anomaly, although in this case, the exposure is intentional by design rather than accidental error leakage. The lack of authorization checks on individual resource accessors is a common pattern in RESTful APIs that have not been rigorously audited for object-level security requirements.
To mitigate this vulnerability, developers must implement strict role-based or attribute-based access control mechanisms within the queryById handler and any other endpoints handling sensitive resources. It is essential to verify that the authenticated user possesses the necessary permissions before returning data related to a specific prompt template. Additionally, the /airag/prompts/list endpoint should be secured to ensure users can only list prompts they are authorized to view or modify. Implementing comprehensive logging for access attempts and monitoring for unusual enumeration patterns can aid in early detection of exploitation activities. Regular security code reviews focusing on authorization logic across all API endpoints are recommended to prevent similar flaws from being introduced during future development cycles.