Executive Summary

A vulnerability in PraisonAI before version 4.6.78 allows unauthenticated attackers to forge message.received events by sending crafted JSON payloads to the webhook endpoint, potentially invoking configured agents with arbitrary sender addresses and message content. This vulnerability has a CVSS score of 8.6 and is classified as HIGH severity. Affected versions are before 4.6.78, and the vulnerability is fixed in version 4.6.78 and later.

Technical Analysis

This vulnerability is a result of PraisonAI's failure to verify Svix webhook signatures in AgentMail webhook mode. The attack vector is NETWORK, and the attack complexity is LOW, requiring no privileges or user interaction. The root cause is the missing input validation and improper access control in the webhook endpoint.

How It Gets Exploited

An unauthenticated remote attacker can send a crafted JSON payload to the webhook endpoint, which is not properly validated due to the lack of Svix webhook signature verification. This allows the attacker to forge message.received events, potentially invoking configured agents with arbitrary sender addresses and message content. The attacker gains the ability to manipulate the message content and sender addresses, which could lead to further exploitation or unauthorized actions within the system.

Impact Assessment

The vulnerability affects PraisonAI versions before 4.6.78. An attacker can achieve a high impact on integrity (I:H) and a low impact on confidentiality (C:L) and availability (A:L). The CVSS score is 8.6, indicating a high severity vulnerability. The blast radius is potentially large, as an attacker could exploit this vulnerability to manipulate messages and potentially pivot to other parts of the system.

Recommended Actions

Update PraisonAI to version 4.6.78 or later to fix this vulnerability. Implement proper input validation and access control measures for the webhook endpoint. Monitor the webhook endpoint for suspicious activity and implement logging and alerting mechanisms to detect potential exploitation attempts.

Sources

  • National Vulnerability Database (NVD)
  • [email protected]
  • https://nvd.nist.gov/vuln/detail/CVE-2026-61436