Proofpoint discloses BluePhishing: OAuth consent phishing abuses Microsoft Entra Agent ID to plant persistent agent identities
Matéria consolidada. Reúne 2 matérias que o Tifon tinha publicado separadamente sobre a mesma história. A IA reescreveu o relatório com o que as 2 fontes trouxeram em 11/11/2026 12:19; os links antigos trazem para cá.
Fontes: gbhackers.com ↗ securityonline.info ↗
Proofpoint researchers disclosed a new cloud intrusion technique called BluePhishing, which abuses the Microsoft Entra Agent ID framework and AI agent model together with OAuth consent phishing (securityonline.info, gbhackers.com). With a single consent click from a victim, an attacker can create persistent agent identities inside the victim tenant. Malicious agent blueprints can generate up to 250 local agent identities, which could widen the attacker's foothold considerably (securityonline.info). Proofpoint describes the work as a proof-of-concept, and no active exploitation or campaigns have been observed in the wild as of 2026-11-11 (gbhackers.com, securityonline.info). One source adds that recent Microsoft 365 configuration changes in November 2026 have increased enterprise exposure to this method (securityonline.info).
Detalhes técnicos
- The attacker configures a malicious blueprint in a tenant they control, then persuades victims to consent to its permissions through OAuth consent phishing (gbhackers.com).
- Consent results in the creation of a blueprint principal in the victim tenant, with persistent permissions (gbhackers.com).
- From there the attacker can request tokens and fabricate autonomous agent behavior. This blends in with normal activity and makes detection difficult (gbhackers.com).
- Per securityonline.info, the blueprint can spawn up to 250 local agent identities inside the target network.
Remediação
Guidance reported by gbhackers.com:
- Monitor specific Azure AD activity logs for agent and blueprint-related activity. The source does not name the log events.
- Apply least-privilege policies.
- Educate users about the risks of third-party agents and consent prompts.
No CVE, patch or IOC was published in the sources.
Linha do tempo
-
BluePhishing Exploits Microsoft Entra OAuth to Create Persistent AI Agent Identities
ler a análise desta fonte ocultar
Researchers have identified a novel attack technique called 'BluePhishing' that leverages Microsoft Entra's AI agent model and OAuth consent phishing. The attacker can, with a single consent click, create persistent agent identities within a victim tenant, potentially expanding the attack surface substantially. Proofpoint notes this is a proof-of-concept and no active exploitation has been observed in the wild. The attack involves configuring a blueprint in a controlled tenant, then convincing victims to consent to its permissions, resulting in the creation of a blueprint principal with persistent, impervious permissions. Once inside, attackers can request tokens and fabricate autonomous agent behavior, blending in with normal activity and making detection challenging. Defenders are advised to monitor specific Azure AD activity logs, implement least privilege policies, and educate users about third-party agent risks.
-
Proofpoint Uncovers BluePhishing Attack Targeting Microsoft Entra Agent Identities
ler a análise desta fonte ocultar
Security researchers at Proofpoint have identified a new cloud intrusion technique called BluePhishing, which exploits the Microsoft Entra Agent ID framework. The attack involves creating malicious agent blueprints that can generate up to 250 local agent identities inside a target network, potentially expanding an attacker's foothold significantly. Although no active campaigns have been observed yet, recent configuration changes in Microsoft 365 in November 2026 have increased enterprise exposure to this method.