Yarden Porat

CrewAI contains a server-side request forgery condition that allows for the acquisition of content from internal and cloud services. This is facilitated by Retrieval-Augmented Generation (RAG) search tools that do not properly validate URLs provided during runtime. The issue allows an attacker to potentially access resources that should not be publicly accessible.

The software contains a flaw where the JSON loader tool reads files without proper path validation. This allows unauthorized access to files on the server. The issue involves an arbitrary local file read.

The CodeInterpreter tool within CrewAI reverts to SandboxPython when Docker is unreachable. This fallback can allow for Remote Code Execution (RCE) through the ability to call arbitrary C functions. The tool utilizes the `SandboxPython` environment, and the issue arises when it cannot connect to Docker, potentially enabling malicious code execution.

CrewAI does not adequately verify the continued operation of Docker during runtime. If Docker is not running, the software reverts to a sandbox configuration that permits Remote Code Execution (RCE) exploitation.

An integer overflow occurs in the `pdf-image.c` file within the `pdf load image imp()` function. This issue allows an attacker to use a specially crafted PDF to trigger the overflow, leading to a heap out-of-bounds write, which may result in arbitrary code execution.

The LDAP authentication method in Vault and Vault Enterprise may not have correctly enforced multi-factor authentication (MFA) when `username as alias` was set to true and a user had multiple Common Names (CNs) that were equal but contained leading or trailing spaces.

Vault and Vault Enterprise’s login MFA rate limits could be bypassed, and TOTP tokens could be reused.

The user lockout feature in Vault and Vault Enterprise could be bypassed for Userpass and LDAP authentication methods.

A privileged Vault operator with write permissions to the root namespace’s identity endpoint could escalate their own or another user’s token privileges to Vault’s root policy.

A privileged Vault operator within the root namespace, possessing write permission to the `{{sys/audit}}` endpoint, may achieve code execution on the underlying host if a plugin directory is configured within Vault’s configuration. The vulnerability allows for remote code execution (RCE) via misconfigured plugin directories.

The Time-based One-Time Password (TOTP) Secrets Engine in Vault and Vault Enterprise is susceptible to code reuse within its validity period due to an issue with the code validation endpoint.

The TLS certificate auth method in Vault and Vault Enterprise did not correctly validate client certificates when configured with a non-CA certificate as a trusted certificate. This allowed an attacker to potentially craft a malicious certificate to impersonate another user.

A timing side channel in the userpass authentication method allowed an attacker to differentiate between existing and non-existing users, potentially enabling the enumeration of valid usernames.

Conjur provides secrets management and application identity for infrastructure. A malformed regular expression allows an attacker manipulating headers signed by AWS to redirect the authentication validation request sent by Secrets Manager, Self-Hosted to a malicious server. This redirection could bypass the IAM Authenticator, granting the attacker permissions granted to the client whose request was manipulated.

An attacker with access to a misconfigured network device routing traffic from Secrets Manager to AWS can redirect authentication requests to a malicious server under their control. CyberArk believes that very few installations are susceptible to active exploitation.

Conjur provides secrets management and application identity for infrastructure. An authenticated attacker who is able to load policy can use the policy YAML parser to reference files on the Secrets Manager, Self-Hosted server. These references may be used for reconnaissance to understand the folder structure of the Secrets Manager/Conjur server or to have the YAML parser include files on the server in the YAML that is processed when the policy loads.

Conjur provides secrets management and application identity for infrastructure. Missing validations in Secrets Manager, Self-Hosted allows authenticated attackers to inject resources into the database and bypass permission checks.

Conjur provides secrets management and application identity for infrastructure. An authenticated attacker who can inject secrets or templates into the Secrets Manager, Self-Hosted database could exploit an exposed API endpoint to execute arbitrary Ruby code within the Secrets Manager process. This issue affects both Secrets Manager, Self-Hosted and Conjur OSS.