DEV Community

CVE Reports
CVE Reports

Posted on Originally published at cvereports.com

CVE-2026-53998: CVE-2026-53998: Format String Injection and Information Disclosure in Arize Phoenix applyChatTemplate

CVE-2026-53998: Format String Injection and Information Disclosure in Arize Phoenix applyChatTemplate

Vulnerability ID: CVE-2026-53998
CVSS Score: 10.0
Published: 2026-11-17

CVE-2026-53998 is a critical format string injection vulnerability in Arize Phoenix affecting versions 13.0.0 prior to 15.7.0. The flaw resides in the applyChatTemplate GraphQL endpoint, which passes user-supplied template strings directly to Python's native str.format() method via FStringTemplateFormatter.format. Because attribute access is not restricted, an attacker can traverse Python dunder attributes (class, __init, __globals_) to read arbitrary process memory and environment variables, including the PHOENIX_SECRET cryptographic key and third-party API credentials.

TL;DR

An unrestricted Python format string evaluation in Arize Phoenix's applyChatTemplate GraphQL endpoint allows unauthenticated attackers to traverse object attributes and extract sensitive environment variables, such as PHOENIX_SECRET, enabling administrative session forgery and credential decryption.


⚠️ Exploit Status: POC

Technical Details

  • Vulnerability Type: Format String Injection / Information Disclosure (CWE-134, CWE-200)
  • Affected Component: FStringTemplateFormatter._format via applyChatTemplate GraphQL endpoint
  • Attack Vector: Network (Unauthenticated HTTP POST to /graphql by default)
  • CVSS v3.1 Score: 10.0 Critical (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:L)
  • EPSS Score: 0.36% (29.86th Percentile)
  • Impact: Disclosure of os.environ (PHOENIX_SECRET, API keys), session forgery, credential decryption
  • Exploit Status: Proof-of-Concept (Regression tests public in patch commit)
  • Fixed Version: arize-phoenix 15.7.0

Affected Systems

  • arize-phoenix
  • arize-phoenix-evals
  • arize-phoenix-otel
  • arize-phoenix-client
  • arize-phoenix: >= 13.0.0, < 15.7.0 (Fixed in: 15.7.0)

Code Analysis

Commit: e7224a6

fix: prevent format-string injection in f-string template formatter (#13197)

class _SafeFieldFormatter(Formatter):
    def get_field(self, field_name: str, args: Any, kwargs: Any) -> Any:
        first = re.match(r"^[^.\[\]]*", field_name).group(0)
        rest = field_name[len(first) :]
        key: Any = int(first) if first.isdigit() else first
        obj = self.get_value(key, args, kwargs)
        for attr, index in _FIELD_NAME_PART_RE.findall(rest):
            if attr:
                if attr.startswith("_"):
                    raise TemplateFormatterError(
                        f"Access to attribute '{attr}' is not permitted in templates"
                    )
                obj = getattr(obj, attr)
            else:
                obj = obj[int(index) if index.lstrip("-").isdigit() else index]
        return obj, first
Enter fullscreen mode Exit fullscreen mode

Exploit Details

  • GitHub: Official patch commit containing regression test payloads demonstrating dunder attribute traversal and os.environ extraction.

Mitigation Strategies

  • Upgrade arize-phoenix to version 15.7.0 or later to enforce attribute filtering via _SafeFieldFormatter.
  • Rotate PHOENIX_SECRET and all environment credentials (LLM API keys, database passwords, cloud keys) if the instance was exposed to untrusted networks.
  • Restrict network access to the Phoenix HTTP/GraphQL interface (port 6006) using firewall policies or private networking.
  • Deploy WAF inspection rules on the /graphql endpoint to block requests containing Python dunder attribute patterns.

Remediation Steps:

  1. Update the package via pip: pip install --upgrade 'arize-phoenix>=15.7.0'
  2. Generate a new cryptographic secret for PHOENIX_SECRET and restart the Phoenix server.
  3. Revoke and reissue third-party API keys (e.g., OPENAI_API_KEY, ANTHROPIC_API_KEY) and database credentials present in the container or host environment.
  4. Verify that PHOENIX_ENABLE_AUTH=True is configured in production environments.

References


Read the full report for CVE-2026-53998 on our website for more details including interactive diagrams and full exploit analysis.

Top comments (0)