CVE-2026-111692: DataBuff ai-apm-web: hardcoded JWT secret enables admin bypass

CRITICALCISA: ATTEND

Published October 16, 2026

CISO Take

DataBuff ai-apm-web through 0.1.10 ships with a hardcoded default HMAC secret (dev-secret) for signing session tokens, so any unauthenticated attacker can forge an admin JWT and gain full control of the /webapi routes, including the Doris SQL query API. The flaw is network-reachable, needs no privileges or user interaction, and is trivial to exploit (CVSS 9.8) because the secret is public in the source repository and a detailed third-party write-up exists. No public exploit, Nuclei template, or KEV listing is recorded yet and EPSS is not available, so exploitation likelihood is unquantified, but the low skill barrier means that could change quickly. Treat any internet-exposed or flat-network DataBuff instance as potentially compromised: no patched version is listed, so immediately rotate the JWT secret to a strong random value via configuration, restrict access to the web API, and review the Doris query logs for unexpected admin activity.

Sources:NVDVulnCheckATLAShackmd.iogithub.com/databufflabs/databuff

What is the risk?

Critical. Exploitation requires only knowledge of the default secret published in the open-source repository (CWE-321, hardcoded cryptographic key), with no authentication, user interaction, or special conditions. Impact covers full confidentiality, integrity, and availability of the APM platform. Exposure depends on whether the web service is reachable and whether operators overrode the default secret. Exploitation status is unconfirmed (not in CISA KEV, no known public exploit, EPSS N/A), but a public technical write-up lowers the barrier. Package adoption and dependents are unknown, which limits blast-radius estimates.

How does the attack unfold?

Reconnaissance

Attacker identifies an exposed DataBuff ai-apm-web instance and reads the default HMAC secret from the public source code.

AML.T0095.000

Initial Access

Attacker signs a forged admin JWT with a future expiry using the dev-secret and presents it to /webapi routes.

AML.T0091.000

Exploitation

Using admin privileges, the attacker runs arbitrary queries via the Doris SQL query API to read stored telemetry.

AML.T0049

Impact

Attacker exfiltrates sensitive telemetry and credentials, and tampers with or deletes monitoring data to hide activity.

AML.T0025

Reconnaissance

Attacker identifies an exposed DataBuff ai-apm-web instance and reads the default HMAC secret from the public source code.

AML.T0095.000

Initial Access

Attacker signs a forged admin JWT with a future expiry using the dev-secret and presents it to /webapi routes.

AML.T0091.000

Exploitation

Using admin privileges, the attacker runs arbitrary queries via the Doris SQL query API to read stored telemetry.

AML.T0049

Impact

Attacker exfiltrates sensitive telemetry and credentials, and tampers with or deletes monitoring data to hide activity.

AML.T0025

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
databuff	—	—	No patch

Do you use databuff? You're affected.

How severe is it?

CVSS 3.1

9.8 / 10

EPSS

0.6%

chance of exploitation in 30 days

Higher than 49% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○CISA SSVC: Public PoC

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVNetwork

ACLow

PRNone

UINone

SUnchanged

CHigh

IHigh

AHigh

What should I do?

1 step

1) No patched version is identified; check the vendor repository (databufflabs/databuff) for a fix newer than 0.1.10 and upgrade as soon as one exists. 2) Immediately override the JWT secret in application.yml (or its environment variable) with a long, random, unique value, then restart the service so all existing tokens are invalidated. 3) Restrict network access to ai-apm-web and /webapi through VPN, IP allowlists, or a reverse proxy with an additional authentication layer; never expose it to the internet. 4) Restrict the Doris database account used by the app to least privilege and limit network reach to Doris. 5) Detect: review web and Doris query logs for admin-session activity without a corresponding login event, tokens with unusual expiry, and unfamiliar source IPs; hunt for SQL queries against the Doris API. 6) If exposure is suspected, rotate all credentials and API keys that may appear in stored telemetry.

What does CISA's SSVC say?

DecisionAttend

Exploitationpoc

AutomatableYes

Technical Impacttotal

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Auth BypassData ExtractionFrameworkAPIAML.T0012- Valid Accounts AML.T0049- Exploit Public-Facing Application AML.T0091.000- Application Access Token

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 15-Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6-AI system operation and monitoring

NIST AI RMF

MANAGE 2.3-Respond to and manage previously unknown risks

OWASP LLM Top 10

LLM03-Supply Chain

Frequently Asked Questions

What is CVE-2026-111692?

Is CVE-2026-111692 actively exploited?

No confirmed active exploitation of CVE-2026-111692 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-111692?

1) No patched version is identified; check the vendor repository (databufflabs/databuff) for a fix newer than 0.1.10 and upgrade as soon as one exists. 2) Immediately override the JWT secret in application.yml (or its environment variable) with a long, random, unique value, then restart the service so all existing tokens are invalidated. 3) Restrict network access to ai-apm-web and /webapi through VPN, IP allowlists, or a reverse proxy with an additional authentication layer; never expose it to the internet. 4) Restrict the Doris database account used by the app to least privilege and limit network reach to Doris. 5) Detect: review web and Doris query logs for admin-session activity without a corresponding login event, tokens with unusual expiry, and unfamiliar source IPs; hunt for SQL queries against the Doris API. 6) If exposure is suspected, rotate all credentials and API keys that may appear in stored telemetry.

What systems are affected by CVE-2026-111692?

This vulnerability affects the following AI/ML architecture patterns: AI observability and APM platforms, Monitoring and telemetry pipelines, LLM application infrastructure, Doris-backed analytics stores.

What is the CVSS score for CVE-2026-111692?

CVE-2026-111692 has a CVSS v3.1 base score of 9.8 (CRITICAL). The EPSS exploitation probability is 0.61%.

What is the AI security impact?

Affected AI Architectures

AI observability and APM platformsMonitoring and telemetry pipelinesLLM application infrastructureDoris-backed analytics stores

MITRE ATLAS Techniques

AML.T0012Valid Accounts

AML.T0049Exploit Public-Facing Application

AML.T0091.000Application Access Token

Compliance Controls Affected

EU AI Act: Art. 15

ISO 42001: A.6.2.6

NIST AI RMF: MANAGE 2.3

OWASP LLM Top 10: LLM03

What are the technical details?

Original Advisory

DataBuff ai-apm-web through 0.1.10 contains an authentication bypass vulnerability that allows unauthenticated attackers to forge admin session tokens by using the hardcoded HMAC secret in JwtTokenService. Attackers can sign a payload of admin and a future expiry with the default dev-secret to gain full administrative access to protected /webapi routes, including the Doris SQL query API.

Exploitation Scenario

An attacker finds an internet-exposed DataBuff ai-apm-web instance via a scan. They read the default dev-secret from the public GitHub source, sign a JWT payload with an admin role and a far-future expiry using HMAC, and send it as the session token to /webapi routes. With admin access they call the Doris SQL query API to dump traces and logs from monitored AI applications, harvesting LLM API keys, prompts, and customer data, then delete or alter records to hide their activity and use the harvested credentials to move into the AI infrastructure.