/

CVE-2026-75111

Severity:

HIGH RISK

(7.5)

Path Traversal (CWE-22) in Evidently UI Materializer

CVE-2026-75111

Deferred

Vulnerability type

Category:

Path Traversal

CWE:

CWE-22

Impact:

Arbitrary File Read and Information Disclosure

Severity:

HIGH RISK

(7.5)

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

Evidently UI does not adequately validate the filename supplied to its dataset materialization endpoint. An unauthenticated attacker may use traversal sequences or absolute paths to read files outside the intended workspace when the materialized dataset is later retrieved.

Affected context

The vulnerability may occur in Evidently UI deployments that expose the dataset materialization functionality and use a client-controlled filename to locate or create dataset content. Risk increases when the service is reachable without authentication and the application process can read sensitive files outside its designated workspace.

Exploit preconditions

  • Attack surface: Dataset materialization and download endpoints in Evidently UI
  • Required access: Network access to the exposed service; authentication may not be required
  • Trigger point: Processing the filename parameter during dataset materialization
  • Why exploitation is possible: The filename is insufficiently constrained to the workspace directory, allowing path resolution outside the intended storage boundary

Why this happens

The underlying bug is improper neutralization of path elements in a user-controlled filename. When an application combines attacker-influenced path components with filesystem operations without enforcing a trusted directory boundary, traversal or absolute path resolution can reach files outside the permitted workspace. This behavior maps to CWE-22: Improper Limitation of a Pathname to a Restricted Directory, commonly called path traversal.

Potential impact

  • An attacker may read sensitive application or system files
  • Credentials, tokens, or configuration data could be disclosed
  • Retrieved datasets may expose internal operational or user information
  • Disclosed secrets could enable follow-on compromise of connected services

General mitigation guidance

Mitigation generally focuses on reducing trust in client-supplied paths and enforcing a filesystem boundary.

Common defensive patterns include:

  • Accepting opaque dataset identifiers rather than arbitrary filenames
  • Resolving candidate paths against a fixed workspace and verifying that the result remains within that directory
  • Rejecting absolute paths, traversal components, and unexpected path separators
  • Applying strict allowlists for permitted filename formats
  • Keeping service credentials and sensitive files outside readable application locations
  • Running the service with the least filesystem permissions required
  • Returning only explicitly authorized materialized datasets through download operations

How teams detect issues like this

Organizations commonly detect path traversal risks through:

  • Static analysis rules for filesystem APIs receiving request parameters
  • Code review checks for path joining without canonicalization and boundary validation
  • Tests using traversal markers, absolute-path forms, encoded separators, and platform-specific path syntax
  • Web application policies that flag suspicious filename parameters and unexpected file access
  • Runtime monitoring for reads outside approved workspace directories

Automated application security testing and code review tools may help identify unsafe path-handling patterns, but findings should be validated against deployment-specific filesystem permissions.

Sources

SourceLink
github.comgithub.com/evidentlyai/evidently
github.comgithub.com/evidentlyai/evidently/blob/main/src/evidently/ui/service/datasets/data_source.py
github.comgithub.com/evidentlyai/evidently/issues/1887
vulncheck.comvulncheck.com/advisories/evidently-ui-path-traversal-via-dataset-materialization-filename

Frequently asked

questions

Yes. When the affected endpoints are network-accessible and unauthenticated, an attacker may be able to trigger the issue remotely.

Not necessarily. The described behavior permits exploitation without authentication, although deployment-specific access controls may limit exposure.

Treat filenames as untrusted input, use controlled identifiers or allowlisted names, and verify that every resolved path remains within the intended workspace.

Combine strict input validation with least-privilege filesystem permissions, boundary-aware path handling, authorization checks, and tests covering traversal and absolute-path variants.

Find out what's already
exploitable in your codebase.