/

CVE-2026-82964

Severity:

HIGH RISK

(8.8)

Permission Escalation (CWE-281) in Avast Sandbox Driver

CVE-2026-82964

Awaiting Analysis

Vulnerability type

Category:

Improper Permission Preservation

CWE:

CWE-281, CWE-653, CWE-862

Impact:

Local Privilege Escalation

Severity:

HIGH RISK

(8.8)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

An improper permission-preservation flaw in the Avast sandbox minifilter driver can leave virtualized files with weaker access controls than their originals. A low-privileged process running inside the sandbox may then alter security descriptors, access protected file copies, and potentially execute code with SYSTEM privileges.

Affected context

The issue arises in Windows systems using the Avast sandbox and its aswSnx.sys minifilter driver. It becomes risky when the driver virtualizes sensitive files, fails to apply the source security descriptor, and permits sandboxed processes to request security changes on the resulting objects. The risk is greater when protected data, such as local account credential stores, is included in the virtualization scope.

Exploit preconditions

  • Attack surface: Local Windows kernel file-system filter and sandbox virtualization paths
  • Required access: Ability to execute a low-privileged process inside the Avast sandbox
  • Trigger point: Creation and subsequent access of a virtualized file or directory
  • Why exploitation is possible: The driver attempts to preserve the original DACL without the required permission to set it, silently ignores the failure, and does not adequately restrict security-descriptor changes for sensitive virtualized objects

Why this happens

The driver opens virtualization targets without the WRITE_DAC access required to apply the original security descriptor. Because the failed permission-preservation operation is discarded, virtualized copies may inherit or retain permissive access controls rather than matching the source object's DACL.

This behavior maps primarily to CWE-281: Improper Preservation of Permissions or Access Controls. Related weaknesses include inadequate isolation of protected objects (CWE-653) and missing authorization enforcement for security-sensitive operations (CWE-862). The absence of an IRP_MJ_SET_SECURITY callback represents an additional defense-in-depth gap, but the central failure is the silently ineffective permission-preservation path.

Potential impact

  • A sandboxed process may alter the security descriptor of a virtualized object
  • An attacker could access protected virtualized files, including credential-related data
  • Local NTLM password hashes may be exposed
  • The attacker could potentially execute code with SYSTEM privileges
  • Sandbox isolation and host security boundaries may be weakened

General mitigation guidance

Mitigation generally focuses on preserving security descriptors reliably and enforcing least privilege throughout file virtualization.

Common defensive patterns include:

  • Open target objects with the minimum rights required to apply their intended security descriptors, including rights needed for DACL preservation
  • Treat permission-application failures as security-relevant errors rather than silently continuing
  • Apply secure-by-default ACLs to virtualized objects when source permissions cannot be reproduced safely
  • Explicitly restrict security-descriptor modification requests for sandboxed processes and sensitive directories
  • Register and enforce security-operation handling for relevant file-system request paths, including security changes
  • Reduce the set of sensitive files and directories exposed to virtualization
  • Run sandboxed workloads with minimal token privileges and isolated service identities

How teams detect issues like this

Organizations can look for this weakness through:

  • Static analysis of file-system filter code for object opens that omit rights needed to preserve security descriptors
  • Review of error-handling paths that ignore failures from ACL, ownership, or security-descriptor operations
  • Code review checks for missing handling of security-related I/O request paths
  • Tests that compare source and virtualized-object DACLs under denied-access conditions
  • Runtime monitoring for sandboxed processes requesting security-descriptor changes or accessing virtualized credential stores
  • Policy enforcement that blocks sandbox contexts from modifying ACLs on protected paths

Automated security analysis tools may help identify incomplete authorization and permission-preservation patterns during development.

Sources

SourceLink
github.comgithub.com/MSNightmare/PrettyPrague
support.avast.comsupport.avast.com/en-us/article/Update-Antivirus
support.avg.comsupport.avg.com/SupportArticleView?l=en&urlname=Update-AVG-Antivirus
support.norton.comsupport.norton.com/sp/en/us/home/current/solutions/v20240108175512167
gendigital.comgendigital.com/us/en/contact-us/security-advisories/
x.comx.com/msnightmare2000/status/2094106041046765752

Frequently asked

questions

No direct remote access is indicated. Exploitation requires local code execution inside the affected sandbox, although a separate remote compromise could potentially provide that starting point.

The attacker needs the ability to run a low-privileged process in the Avast sandbox and reach a file virtualization path. Administrative privileges are not required for the initial execution context.

Preserve source permissions using correctly authorized object handles, fail closed when security-descriptor application fails, and explicitly deny unauthorized security changes to virtualized sensitive objects.

Combine least-privilege sandbox tokens, strict virtualization scope, complete handling of security-related file-system requests, negative permission tests, and monitoring for ACL changes from sandboxed processes.

Find out what's already
exploitable in your codebase.