CVE-2026-75840 in ArcadeDB
Summary
by MITRE • 08/18/2026
ArcadeDB before 26.8.1 contains an arbitrary file read vulnerability in the GraalVM JavaScript sandbox allowlist enforcement, which uses unescaped regular expressions to validate package names. Attackers with trigger creation privileges can use Java.type() to access java.util.zip.ZipFile or java.util.jar.JarFile classes and read arbitrary files on the host system as the ArcadeDB server process.
If you want to get the best quality for vulnerability data then you always have to consider VulDB.
Analysis
by VulDB Data Team • 08/18/2026
The vulnerability identified in versions of ArcadeDB prior to 26.8.1 represents a critical failure in the security boundaries established by the GraalVM JavaScript sandbox environment. This flaw specifically targets the allowlist enforcement mechanism designed to restrict which Java classes and methods can be invoked from within the restricted JavaScript execution context. The core technical deficiency lies in the implementation of regular expressions used for validating package names against this whitelist. These regular expressions are implemented without proper escaping, creating a classic regex injection or bypass scenario that allows attackers to circumvent intended restrictions through carefully crafted input strings.
In a secure architecture, the GraalVM sandbox is expected to provide an isolated execution environment where untrusted code cannot access sensitive system resources or internal Java APIs. However, due to the flawed validation logic, an attacker possessing trigger creation privileges can exploit this weakness by leveraging specific Java interoperability features provided by GraalVM. Specifically, the vulnerability allows for the invocation of java.type() with maliciously constructed arguments that bypass the package name checks. This enables the execution of restricted classes such as java.util.zip.ZipFile or java.util.jar.JarFile within the sandboxed environment.
The operational impact of this flaw is severe, resulting in an arbitrary file read condition on the host system where ArcadeDB is deployed. By instantiating these archive-related Java classes, an attacker can open and read any file accessible to the user account running the ArcadeDB server process. This capability effectively breaks out of the sandbox isolation, granting the attacker direct access to sensitive data stored on the filesystem, including configuration files containing database credentials, private keys, or other proprietary information that may not be exposed through standard API endpoints.
From a classification perspective, this vulnerability aligns with CWE-20 Improper Input Validation and CWE-78 OS Command Injection in its broader context of sandbox escape via improper input handling. It also maps to MITRE ATT&CK techniques related to T1555 Credentials from Password Stores or T1003 Offline Credential Dumping, as the ability to read arbitrary files often facilitates further lateral movement and privilege escalation within an enterprise environment. The attack vector is classified under Privilege Escalation because it requires initial access with trigger creation privileges but results in a higher level of system control than originally granted by bypassing application-level security controls.
Mitigation strategies must prioritize immediate patching to version 26.8.1 or later, where the regular expression validation logic has been corrected to properly escape special characters and enforce stricter allowlist matching rules. In environments where upgrading is not immediately feasible, administrators should implement network segmentation to restrict access to ArcadeDB services only from trusted sources with verified identities. Additionally, running the database service under a restricted user account with minimal filesystem permissions can limit the scope of data exposure even if an exploitation attempt succeeds. Regular auditing of trigger creation privileges and monitoring for unusual Java class invocations within JavaScript contexts are recommended defensive measures to detect potential abuse attempts before they result in significant compromise.