Prisma Access Agent CVE-2026-0268

Exploitation requires local access to a Linux endpoint with Prisma Access Agent installed and an active or configured VPN session, with the attacker holding at minimum a standard (low-privileged) user account on the system. The vulnerability is Linux-platform-specific - the same agent software on Windows, macOS, iOS, Android, and ChromeOS is not affected, implying the bypass depends on Linux-specific networking or agent implementation behavior. No special non-default configuration is required based on the AT:N (no attack requirements) in the CVSS 4.0 vector, suggesting this is exploitable against standard Prisma Access Agent Linux deployments. Exploitation is strictly limited by the requirement for prior local access; remote unauthenticated exploitation is not possible. The E:U exploit maturity and AU:N (not automatable) supplemental ratings confirm this is not suitable for mass exploitation and is constrained to targeted insider or post-compromise scenarios.

The CVSS 4.0 score of 4.4 (Medium) is appropriate given the local-only attack vector and low-privilege requirement. AV:L and PR:L together mean this is not remotely exploitable - an attacker must already have a foothold on the targeted Linux endpoint, making this primarily an insider threat or post-compromise escalation scenario. The confidentiality impact VC:H is justified because bypassing the VPN tunnel can expose sensitive traffic - internal API calls, credentials, proprietary data - to interception on untrusted networks, defeating the primary purpose of the VPN deployment. The supplemental CVSS 4.0 metrics provide additional signal: E:U (Unexploited) confirms no known active exploitation, AU:N (not automatable) means mass exploitation is not feasible, and the Urgency rating of Amber reflects a real but non-critical priority. The vulnerability is absent from the CISA KEV catalog, consistent with E:U. EPSS data was not included in the provided intelligence and cannot be assessed. Organizations running Linux-based Prisma Access Agent deployments with high-sensitivity workloads - particularly where VPN bypass would evade DLP, logging, or network threat inspection - face the highest material risk.

A low-privileged user with a local account on a Linux endpoint running Prisma Access Agent manipulates the system's routing state or interacts with the agent's enforcement mechanism to cause network traffic to egress outside the VPN tunnel. The attacker then transmits or intercepts sensitive data - such as internal service credentials or proprietary communications - over an untrusted network path, bypassing VPN-enforced DLP, packet inspection, and audit logging controls. No public proof-of-concept exists at time of analysis, so exploitation likely requires specific knowledge of the bypass mechanism and direct local system access.