Status: Reported to Google VRP (Report #450403279) on 2025‑10‑09. Vendor response: classified as “Working as Intended” / social engineering.
This repository hosts a responsibly redacted write‑up intended for defensive research and vendor communication.
Google Translate’s site-proxying flow can render third‑party pages under a *.translate.goog origin. In a tested PoC, attacker‑controlled JavaScript executed under that Google‑managed origin and performed silent exfiltration of client metadata (IP, UA/UAData, device metrics). This repository contains sanitized, non‑actionable material only — no runnable PoC, raw collector URLs, or unredacted PII.
The issue was reported to Google VRP (Report #450403279). After review, the vendor responded:
"The issue you're describing can only result in social engineering. We don't consider this type of open redirect to be a vulnerability."
Key points:
- Google acknowledges that the technical behavior exists — the exfiltration of client metadata via the tested PoC is valid and reproducible.
- The vulnerability was previously categorized as an open redirect, because the PoC appeared to only redirect a site under the Google domain. However, the PoC demonstrates automatic data exfiltration that occurs within a Google-managed sandbox domain, which technically qualifies this issue as a sandbox escape.
- Their classification as “social engineering / not a vulnerability” reflects a policy or risk-assessment decision rather than a dispute of the technical findings.
- The repository documents the technical behavior and sanitized PoC for defensive awareness and research purposes only.
- This demonstrates that even when a vendor classifies an issue as low-risk, the underlying behavior can have real security implications that are valuable for defensive study.
Note on Official Report:
The official report submitted to Google VRP, including the full PoC, raw collector endpoints, and other confidential details, is private and cannot be shared publicly. This repository contains only sanitized, non-actionable material for defensive review and research purposes. Access to the unredacted report is restricted and may be provided only under NDA to verified vendor contacts or CERT teams.
This repository contains a layered, redacted analysis of a reported vulnerability in Google Translate. To navigate it effectively:
-
Start with the TL;DR
Provides a high-level summary of the findings and the nature of the vulnerability without exposing actionable exploit details. -
Review the ATTACK_SCENARIO.md
Understand the logical flow of the reported issue, sanitized for safe review. Focus on methodology, bypass concept, and impact, not specific payloads. -
Read TECHNICAL_ANALYSIS.md
Offers a deeper technical perspective on the sanitization failure and the sandbox containment model. All sensitive details are redacted; the emphasis is on concepts and risk understanding. -
Consult TIMELINE.md and ADJUDICATION_VRP.md
Learn the disclosure and vendor response process. Observe how the report progressed, how VRP classifications were applied, and the reasoning behind final status decisions. -
Examine SANITIZED_ARTIFACTS/
Contains screenshots and redacted payload examples. These illustrate the observed behavior in a safe, non-actionable form. -
Read DISCLOSURE.md and FAQ.md
Offers guidance on responsible disclosure, research methodology, and answers common questions for defenders or other researchers. -
Focus on Concepts, Not Reproduction
All executable content, exact payloads, collector URLs, and PII have been removed. The goal is to understand the security implications and mitigation strategies, not to replicate the exploit.
Following this reading order ensures a clear understanding of the findings, their significance, and defensive recommendations without exposing sensitive technical details.
- ATTACK_SCENARIO.md — Non‑actionable attack scenario (sanitized).
- TECHNICAL_ANALYSIS.md — Redacted low‑level analysis of the observed Content Sanitization Bypass (CSB).
- TIMELINE.md — Chronological disclosure and vendor response timeline.
- ADJUDICATION_VRP.md — Notes on the VRP adjudication and policy considerations.
- DISCLOSURE.md — Record of communications and escalation.
- SANITIZED_ARTIFACTS/ — Screenshots and sanitized artifacts showing origin and exfiltration evidence (redacted).
- FAQ.md — Frequently asked questions for defenders and triage teams.
- README.md — This file.
This repository intentionally omits details that would enable reproduction of the exploit. Specifically, any runnable payloads, exact injection vectors, raw collector endpoints, and full unredacted logs are not published here.
- Raw, unredacted evidence is archived offline and integrity‑hashed.
- Unredacted artifacts may be shared only with verified vendor contacts, CERTs, or under a mutual Non‑Disclosure Agreement (NDA).
This repository is intended for informational and defensive purposes only. It documents findings and the disclosure process; it is not an exploit repository. Use responsibly.
This repository contains original analysis and redacted artifacts. Redistribution of unredacted material is prohibited except under explicit agreement (e.g., NDA).