TrendAI Vision One: Log Information Disclosure Fixed After 10 Months

On July 29, 2026, advisory ZDI-26-497 publicly disclosed an information disclosure vulnerability in the TrendAI Vision One Service Gateway module. The flaw had been fixed months earlier but remained hidden until the end of a coordinated disclosure process lasting ten months. Cataloged as CVE-2025-71386 with a CVSS 3.1 score of 7.7, the vulnerability allowed an authenticated remote attacker to extract sensitive information from log files — a paradox for an enterprise cybersecurity platform.

The Mechanism: When Logs Become a Goldmine

The specific flaw, documented in advisory ZDI-26-497, is located in the TrendAI Vision One Service Gateway. The component wrote sensitive information directly to a log file, turning a diagnostic artifact into an attack surface. The exact nature of this sensitive data is not detailed in either the ZDI advisory or the vendor bulletin, which limit themselves to the generic classification of "information disclosure."

Authentication is a prerequisite for exploitation. The attacker must possess valid credentials, which narrows the risk perimeter but does not eliminate it: compromised accounts, malicious insiders, or escalation from related vulnerabilities can provide the entry point. The remote attack vector (AV:N) and low attack complexity (AC:L) indicate that, once credentials are obtained, the technical barrier is minimal.

"The specific flaw exists within the Service Gateway module. The issue results from storing sensitive information in a log file. An attacker can leverage this vulnerability to disclose sensitive information, leading to further compromise." — Advisory ZDI-26-497

The impact scope includes a scope change (S:C), a technically significant detail: the vulnerability can extend its effects beyond the compromised user's authorized resources, touching other components or data in the same infrastructure. This is consistent with the architecture of a managed multi-tenant security platform, where system logs often aggregate streams from multiple customers.

The Timeline of a Two-Phase Closure

TrendAI's response unfolded in distinct phases, documented in bulletin KA-0023937. The critical part of the vulnerability was mitigated on the backend in September 2025, immediately after researcher Hugo LECLERCQ reported it. This rapid intervention reduced operational risk without requiring customer action.

However, "remnant" material — the term "remnant updates" used in the bulletin — required two monthly service update cycles: September and October 2025. The distinction between backend mitigation and client-side updates suggests an architectural separation: the vulnerable logic could be fixed centrally, but pre-existing logs or agent-side configurations required distributed deployment. The entire process concluded by December 2025, when the bulletin declares both vulnerabilities in the document (CVE-2025-71386 and the related CVE-2025-71387) resolved.

The time window between effective mitigation and public disclosure — approximately ten months — is the price of responsible coordination. On September 17, 2025, the researcher submitted the report; on July 29, 2026, ZDI published the advisory. This latency is standard in disclosure protocol, but it creates an information asymmetry: regularly updated customers were protected without knowing it, while unmaintained installations could have remained exposed until publication, unaware the threat even existed.

Security Platforms and the Secure Logging Problem

The case reveals a structural tension in the enterprise cybersecurity sector. Security platforms — in this case an AI-native solution for visibility and incident response — must log diagnostic details to function, but that same logging abundance becomes a risk when logs contain sensitive materials. It is an instance of CWE-532 (Insertion of Sensitive Information into Log File), a weakness category the MITRE CWE project recognizes as recurring but underestimated.

The vendor bulletin's "information only" nature, with the explicit statement that "no action is required for customers who update regularly," is a communication choice that minimizes alarm. It is technically correct — the vulnerability is already fixed — but it reduces incident visibility for decision-makers who do not monitor service updates with the same attention as critical patches. The distinction between "no action required" and "no historical impact" is subtle but relevant: the dossier does not document whether the exposed logs were actually accessed by threat actors before the September 2025 mitigation.

The CVSS 3.1 score of 7.7, with zero impact on integrity and availability (I:N, A:N), classifies the flaw as high severity but not critical. The priority is confidentiality, in a product whose primary function is precisely to protect the confidentiality of others' data. The reputational risk outweighs the immediate technical risk here: the news that a security platform fails to protect its own logs reinforces the narrative of a fragile supply chain even among the guardians.

What to Do Now

For TrendAI Vision One installations, the vendor bulletin is clear: the monthly updates distributed by December 2025 resolved the vulnerability. Verifying the Service Gateway status and installed client versions is the only necessary check. The absence of known public exploits and the closure of the opportunity window make an active residual risk unlikely.

For the broader sector, the case suggests three checks to incorporate into security platform development processes: reviewing log content before permanent writing, separating system logs from application logs containing user data, and automating sensitive pattern scanning in CI/CD pipelines. These are not recommendations derived from the specific dossier, but standard secure logging practices that the flaw documents as unimplemented in this component.

The source does not specify corrective measures beyond the monthly updates already distributed. The dossier does not document whether TrendAI has modified its development practices or introduced additional controls to prevent similar recurrences. The exact nature of the exposed data — tokens, credentials, configurations, or other — remains undeclared, limiting the historical impact assessment for organizations that may have had compromised logs before the September 2025 mitigation.

The Lesson of the Hole in the Logs of Those Who Watch Others

The vulnerability CVE-2025-71386 is not a spectacular exploit. It is a software engineering error, faulty logging in a central module of an enterprise platform. Its importance derives from context: when the product that sells security exposes its own logs, trust cracks on two levels, technical and perceived.

The ten-month coordinated disclosure worked as intended, but it also left a long gray zone where the vulnerability existed without public acknowledgment. For diligent customers, protection was available; for those who do not update automatically, awareness arrived late. The lesson is not that TrendAI Vision One is less secure than competing alternatives — the dossier does not allow comparisons — but that cybersecurity is a field where even the guardians must be watched, and where logs, often overlooked as a risk surface, can become the richest mine for an already authenticated attacker.

Sources

• http://www.zerodayinitiative.com/advisories/ZDI-26-497/
• http://www.zerodayinitiative.com/advisories/published/
• https://www.cve.org/CVERecord?id=CVE-2025-71386
• http://www.zerodayinitiative.com/advisories/upcoming/
• https://www.trendmicro.com/
• https://success.trendmicro.com/en-US/solution/KA-0023937

Information verified against cited sources and current as of publication.

Sources