closedclosed 2 Dec 2020Research1 Research / proof of conceptAI
Model Extraction and Adversarial Attacks on Translation APIs
Google Translate, Bing Translator, Systran Translate
Researchers replicated commercial machine translation services via public APIs and transferred adversarial inputs to trigger targeted errors. The demonstration proved that black-box models are vulnerable to extraction and evasion.
- first report
- 18 Nov 2020
- last update
- 18 Nov 2020
Timeline · 4
- 18 Nov 2020Attack on Machine Translation Services
- 18 Nov 2020Wallace, Eric, et al. "Imitation Attacks and Defenses for Black-box Machine Translation Systems" EMNLP 2020
- 18 Nov 2020Project Page, "Imitation Attacks and Defenses for Black-box Machine Translation Systems"
- 18 Nov 2020Google under fire for mistranslating Chinese amid Hong Kong protests
Sources · 4
News and write-ups
- 2020-11-18Project Page, "Imitation Attacks and Defenses for Black-box Machine Translation Systems"ericswallace.com · ericswallace.com/imitation
- 2020-11-18Google under fire for mistranslating Chinese amid Hong Kong proteststhehill.com · thehill.com/policy/international/asia-pacific/449164-google-under-fire-for-mistr
Research and catalogues
- 2020-11-18Attack on Machine Translation ServicesMITRE ATLAS · atlas.mitre.org/studies/AML.CS0005
- 2020-11-18Wallace, Eric, et al. "Imitation Attacks and Defenses for Black-box Machine Translation Systems" EMNLP 2020arXiv · arxiv.org/abs/2004.15015
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Research
- AI role
- Attack on AI
- categories
- Model theft / extractionAdversarial evasion
- severity
- 1 · Research / proof of concept
- targets
- Google, Microsoft
- owasp llm top 10
- LLM10:2025 Unbounded Consumption
- mitre atlas
- AML.T0015 Evade AI ModelAML.T0043 Craft Adversarial DataAML.T0048.004 External Harms: AI Intellectual Property Theft
- dates
- happened 30 Apr 2020 · first reported 18 Nov 2020
- origin
- MITRE ATLASDerived from MITRE ATLAS (Apache 2.0, © MITRE).
- record
- confidence 0.95 · created 2026-10-18 06:55Z · updated 2026-10-18 06:55Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# Model Extraction and Adversarial Attacks on Translation APIs https://gcdradar.vercel.app/incidents/model-extraction-and-adversarial-attacks-on-translation-apis · AIR-2020-0029 · closed 2020-12-02 - Lane: Research - Severity: 1 (Research / proof of concept) - Fix status: Fix status unknown - First reported: 2020-11-18 (happened 2020-04-30) - AI role: Attack on AI - Products: Google Translate, Bing Translator, Systran Translate ## Summary Researchers replicated commercial machine translation services via public APIs and transferred adversarial inputs to trigger targeted errors. The demonstration proved that black-box models are vulnerable to extraction and evasion. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2020-11-18: Google under fire for mistranslating Chinese amid Hong Kong protests (https://thehill.com/policy/international/asia-pacific/449164-google-under-fire-for-mistranslating-chinese-amid-hong-kong) - 2020-11-18: Project Page, "Imitation Attacks and Defenses for Black-box Machine Translation Systems" (https://ericswallace.com/imitation) - 2020-11-18: Wallace, Eric, et al. "Imitation Attacks and Defenses for Black-box Machine Translation Systems" EMNLP 2020 (https://arxiv.org/abs/2004.15015) - 2020-11-18: Attack on Machine Translation Services (https://atlas.mitre.org/studies/AML.CS0005) ## Sources - 2020-11-18 MITRE ATLAS: Attack on Machine Translation Services https://atlas.mitre.org/studies/AML.CS0005 - 2020-11-18 arXiv: Wallace, Eric, et al. "Imitation Attacks and Defenses for Black-box Machine Translation Systems" EMNLP 2020 https://arxiv.org/abs/2004.15015 - 2020-11-18 ericswallace.com: Project Page, "Imitation Attacks and Defenses for Black-box Machine Translation Systems" https://ericswallace.com/imitation - 2020-11-18 thehill.com: Google under fire for mistranslating Chinese amid Hong Kong protests https://thehill.com/policy/international/asia-pacific/449164-google-under-fire-for-mistranslating-chinese-amid-hong-kong Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closedEdge AI Product Microsoft Edge AI Product Evasion Red Team Exercise
Research · since Feb 2020 (approx.)
AI - closedAzure Microsoft Internal Azure AI Service Red Team Exercise
Research · since Jan 2020 (approx.)
AI - closedDGA detector Botnet DGA Detection Evasion Using Domain Mutation
Research · since 1 Jan 2020
AI - closedPalo Alto Networks malware detection system Evasion of Deep Learning Detector for Malware C&C Traffic
Research · since 1 Jan 2020
AI - closedCylancePROTECT Bypassing Cylance AI Malware Detection via Universal String
Research · since 7 Sep 2019
AI - closedShanghai government tax office facial recognition Camera Hijack and Facial Recognition Evasion for Tax Fraud
Patch now · Real loss · since Jan 2020 (approx.)
AI