Android CVE-2025-48652

Attacker must already be executing code locally on the target Android 15, 16, or 16-qpr2 device with low privileges (PR:L per CVSS) - for example, a previously installed app or a process reachable via ADB. The device must be subject to MDM policy enforcement for the bypass to have meaningful impact, since the bug specifically defeats the InstallRepository.kt performPreInstallChecks MDM-policy gate during a package install operation. No user interaction is required (UI:N), and exploitation does not depend on a specific OEM customization beyond running an affected AOSP version; unmanaged consumer devices are technically vulnerable but gain little since no MDM policy is being circumvented.

The CVSS 7.8 vector (AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) indicates a local-attack-vector, low-complexity, low-privileged-attacker exploit with no user interaction and high CIA impact - consistent with an on-device privilege escalation rather than a remote compromise. EPSS is 0.01% (1st percentile) and the CVE is not in CISA KEV, so there is no public exploit identified at time of analysis and no evidence of mass exploitation; this aligns with the local-only attack surface, which limits opportunistic scanning. Real-world priority is elevated specifically for enterprise/BYOD fleets that rely on MDM policy enforcement to constrain app installation, since the bug directly undermines that trust boundary - for unmanaged consumer devices the practical risk is lower. Note a minor data conflict to verify: the description states 'local escalation of privilege with no additional execution privileges needed' which reads as PR:N, while the CVSS vector lists PR:L - defenders should treat the lower of the two (PR:L) as a hard requirement only until Google clarifies.

A malicious app already running on a managed device (sideloaded during a brief policy gap, or delivered through a benign-looking app store entry) invokes the package install flow and triggers the flawed performPreInstallChecks path to install a second, higher-privileged or policy-prohibited application that MDM should have blocked. The newly installed app then operates outside MDM restrictions, gaining capabilities such as enterprise data access or removal of management agents. No public exploit identified at time of analysis, and no user interaction is required once the initial local code is running.