/

CVE-2026-93594

Severity:

HIGH RISK

(8.1)

Authorization Bypass (CWE-863) in ArcadeDB Database Engine

CVE-2026-93594

Deferred

Vulnerability type

Category:

Authorization Bypass

CWE:

CWE-863

Impact:

Unauthorized data disclosure and integrity violations

Severity:

HIGH RISK

(8.1)

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 authorization enforcement gap in ArcadeDB allows authenticated users to access indexed and time-series data belonging to types they are not permitted to read or modify. A single ordinary SQL statement may expose indexed keys, record identifiers, aggregates, or sample counts, and may delete index entries, potentially causing data corruption and constraint failures.

Affected context

The issue occurs in ArcadeDB deployments where per-type or per-record access-control rules are expected to govern query access. Query paths involving LSM index files or the TimeSeries engine can bypass checks performed by LocalBucket, while schema index metadata remains broadly discoverable. The behavior may affect both embedded and server deployments across supported transports after an authenticated principal has been established.

Exploit preconditions

  • Attack surface: SQL query execution through an embedded or server deployment
  • Required access: An authenticated account with insufficient read or delete permissions for the targeted type
  • Trigger point: Query operations that access index values, index-derived aggregates, TimeSeries samples, or index entries
  • Additional condition: The relevant type and index names must be discoverable or otherwise known
  • Why exploitation is possible: Some query execution paths do not invoke the per-type or per-record authorization check enforced by LocalBucket

Why this happens

The underlying flaw is inconsistent authorization enforcement across storage and query-engine paths. Access rules are applied in LocalBucket using file identifiers, but equivalent operations routed through LSM indexes or the TimeSeries engine can return or modify data without performing the same permission check. This maps to CWE-863: Incorrect Authorization, because authenticated users can perform actions outside the permissions assigned to their principal.

Potential impact

  • Unauthorized disclosure of indexed key values and record identifiers
  • Exposure of aggregate values, record counts, or TimeSeries samples
  • Deletion of index entries by users lacking delete authorization
  • Desynchronization between indexes and stored records
  • Circumvention or disruption of unique constraints
  • Potential data integrity loss and operational outages

General mitigation guidance

Mitigation generally focuses on making authorization consistent across every query and storage access path.

Common defensive patterns include:

  • Enforcing type- and record-level authorization before index, aggregate, and TimeSeries operations
  • Applying centralized policy checks rather than relying on a single storage-layer implementation
  • Treating schema metadata, index contents, and derived values as protected resources
  • Denying modification of index structures unless the principal is authorized for the underlying data
  • Running database services with least-privilege roles and narrowly scoped permissions
  • Adding authorization regression tests for every supported transport and query execution engine
  • Monitoring and validating index consistency after denied or unauthorized operations

How teams detect issues like this

Detection can combine static review, authorization-focused testing, and runtime monitoring:

  • Review query planners and storage adapters for paths that return or mutate records without centralized authorization checks
  • Search code for permission enforcement implemented only in a single storage class or keyed exclusively by file identifiers
  • Test equivalent operations through local, server, HTTP, Bolt, Postgres, and Gremlin interfaces using restricted principals
  • Verify that schema metadata does not expose protected type or index details beyond policy requirements
  • Add policy tests covering indexed reads, aggregate shortcuts, TimeSeries access, record counts, and index mutations
  • Alert on index-deletion operations or index/data consistency changes performed by low-privilege accounts
  • Use automated security review tools to flag authorization decisions that are absent from alternate execution paths

Sources

SourceLink
github.comgithub.com/ArcadeData/arcadedb/security/advisories/GHSA-2c8m-q484-jv7m
vulncheck.comvulncheck.com/advisories/arcadedb-before-26.9.1-acl-bypass-via-index-and-timeseries

Frequently asked

questions

An authenticated user may be able to exploit it when their assigned permissions deny access to a type but the query reaches an LSM index or TimeSeries execution path that omits the corresponding authorization check.

No. Administrative privileges are not the general prerequisite; the key condition is an authenticated principal with access to query execution but insufficient authorization for the targeted data.

Enforce the same centralized authorization decision before every operation that reads or modifies records, indexes, aggregates, metadata, or TimeSeries data.

Use deny-by-default policies, protect derived and metadata views, test every transport and query path with restricted accounts, and continuously monitor for unauthorized index or data mutations.

Find out what's already
exploitable in your codebase.