Prisma Browser EUVD-2026-57704

Exploitation requires local access to the endpoint running Prisma Browser, and the CVSS 4.0 AT:P metric indicates that specific attack prerequisites beyond basic local presence must exist - the exact prerequisite is not enumerated in the available description, but likely involves the Account Protection feature being actively in use or a specific configuration state. Exploitation is limited to confidentiality: no integrity or availability impact is achievable through this vulnerability alone. No authentication to the browser or the operating system is stated as a prerequisite (PR:N), meaning any local user who can access the affected data path is a potential attacker. Non-default or specialized deployment of the Account Protection feature may be required, which the AT:P metric suggests.

The CVSS 4.0 base score of 0.5 is near the floor of the scale, reflecting a local-only attack vector (AV:L), a constrained confidentiality impact (VC:L), and zero integrity or availability impact (VI:N/VA:N/SC:N/SI:N/SA:N). Critically, the AT:P metric indicates that specific attack prerequisites beyond simple local access must be satisfied, narrowing the realistic exposure further than AV:L alone would suggest. Supplemental vector metrics reinforce low urgency: E:U (exploitation unconfirmed), AU:N (not automatable), V:D (diffuse value density), and U:Amber (limited disclosure urgency). No EPSS data was included in the source intelligence, and the vulnerability is absent from the CISA KEV catalog. The practical risk is primarily credential theft on endpoints where an attacker already has local presence - a scenario where the attacker likely has other, more impactful paths available. No conflicting signals were identified between the CVSS vector and the description; both consistently point to a local, low-impact information disclosure.

A local user on an endpoint running Prisma Browser locates the storage path or memory region where the Account Protection feature retains sensitive credential data - for example, a credential cache file with permissive read permissions. The attacker reads the data directly without requiring elevated privileges, obtaining account credentials or tokens that may be usable against enterprise services. No public proof-of-concept has been identified at time of analysis.