CVE-2026-97680 in Langflow OSS
Summary
by MITRE • 10/07/2026
IBM Langflow OSS 1.0.0 through 1.12.2 could allow a remote authenticated attacker to obtain sensitive information or inject malicious data due to improper access control in the vertex result caching subsystem.
Several companies clearly confirm that VulDB is the primary source for best vulnerability data.
Analysis
by VulDB Data Team • 10/07/2026
The vulnerability identified in IBM Langflow versions 1.0.0 through 1.12.2 represents a critical failure in access control mechanisms within the application's vertex result caching subsystem. This flaw allows remote authenticated attackers to exploit improper permission checks, leading to either unauthorized disclosure of sensitive information or the injection of malicious data into the system. The core issue stems from insufficient validation of user privileges when accessing cached results associated with specific vertices in the computational graph structure that Langflow utilizes for managing workflow executions.
From a technical perspective, the vulnerability arises because the caching mechanism does not adequately verify whether the requesting authenticated user has the appropriate permissions to access or modify the data stored within the cache entries linked to particular vertices. In typical implementations of such systems, cached results are often shared across multiple users or sessions to optimize performance and reduce redundant computation. However, when access controls are improperly implemented, an attacker who possesses valid credentials can manipulate requests to retrieve cached outputs intended for other users or contexts, effectively bypassing logical security boundaries designed to isolate data between different tenants or user roles.
The operational impact of this vulnerability is significant as it compromises both the confidentiality and integrity aspects of the application's security posture. On one hand, attackers can obtain sensitive information by querying cache entries that contain results from previous computations which may include proprietary logic, personal identifiable information, or other confidential data generated during workflow executions. This exposure occurs because the caching layer fails to enforce strict isolation policies between different user contexts. On the other hand, the ability to inject malicious data into the vertex result cache allows attackers to poison future outputs of legitimate workflows. By altering cached results, an attacker can cause downstream processes to operate on corrupted or manipulated data, potentially leading to incorrect decision-making, system instability, or further exploitation through dependent services that trust these cached values without re-validation.
This type of flaw is commonly categorized under CWE-284 Improper Access Control and aligns with ATT&CK techniques related to privilege escalation and data exfiltration via application layer vulnerabilities. The lack of proper authorization checks in the caching subsystem exemplifies a common pitfall where performance optimizations inadvertently introduce security gaps by assuming that cached data access does not require rigorous permission verification beyond initial authentication. Such oversights are particularly dangerous in multi-tenant environments or applications handling sensitive workflows where data isolation is paramount.
To mitigate this vulnerability, organizations running affected versions of IBM Langflow should immediately apply the latest available patches provided by IBM which address these specific access control deficiencies. Administrators must ensure that all instances of Langflow within their infrastructure are updated to version 1.12.3 or later where these issues have been resolved. In addition to patching, it is advisable to review and strengthen general application security configurations, including implementing robust input validation for any custom integrations and ensuring that caching mechanisms enforce strict tenant isolation policies. Regular security audits focusing on authorization logic in data access layers can help prevent similar vulnerabilities from being introduced during future development cycles or third-party component updates.