Fast Facts
- Malicious actors exploit non-human identities (NHIs) to bypass authentication and execute unauthorized workflows without needing user credentials, leading to potential data breaches and system manipulation.
- AI workflows often operate with privileged credentials and lack proper authorization checks, enabling attackers to trigger sensitive actions or access data impersonating trusted agents.
- The core vulnerability is inadequate access governance; workflows share privileged credentials, making it difficult to identify or prevent malicious requests, thus creating a significant attack surface.
Threats, Attack Techniques, and Targets
Researchers from Noma Security found a new backdoor in AI workflows called Workflow Identity Hijacking. This attack exploits non-human identities (NHIs), such as service accounts or API keys. The attackers send benign requests through unauthenticated entry points like web forms, support inboxes, or shared documents. Instead of tricking the AI model, attackers bypass standard controls by using these unauthenticated methods. The AI processes the request exactly as it was designed to, fulfilling it without verifying if the requester has proper permissions. The targets are the production systems and data accessed by these NHIs. Attackers aim to gain unauthorized access to sensitive resources by disguising their requests as legitimate workflow actions.
Impact, Security Implications, and Remediation Guidance
This attack creates serious security risks. It can lead to data breaches or unauthorized actions within systems. Since the attack uses legitimate workflows and identities, traditional defenses like filtering or monitoring for anomalies may not detect it. Security leaders warn that static protections are no longer enough. Instead, teams need real-time governance and continuous visibility of non-human identities. Proper controls include short-lived tokens, explicit authorization checks, and mapping each workflow’s access permissions. As for remediation, it is recommended to consult the relevant vendor or security authority for specific guidance. The best approach is to implement strict identity governance for AI workflows by limiting permissions, adding authorization checkpoints, and maintaining detailed audit logs.
Discover More Technology Insights
Dive deeper into the world of Cryptocurrency and its impact on global finance.
Explore past and present digital transformations on the Internet Archive.
ThreatIntel-V1
