Microsoft Celeb (MS-Celeb-1M) facial recognition dataset

MS-Celeb-1M (or Microsoft Celeb) is a dataset developed by Microsoft Research to accelerate research into facial recognition technologies. 

Created and published in 2016, MS-Celeb-10 consisted of approximately 10 million facial images of 100,000 celebrities, journalists, artists, musicians, activists, policy makers, writers, and academics. Micosoft also provided a 'target list' of an additional 900,000 names whose images were to be collected.

According to Microsoft, the dataset was created for 'non-commercial research purpose only' and would be applicable to image captioning and news video analysis. 

Reckoned to be the largest public dataset of its kind, Microsoft terminated the project mid-2019 shortly after the publication of researcher Adam Harvey's Exposing.ai project and a Financial Times investigation into facial recognition data sharing.

Risks and harms 🛑

Microsoft Celeb is seen to have posed significant privacy and ethical concerns due to its large-scale collection of celebrity images without explicit consent, potentially enabling unauthorised surveillance, identity theft, and the misuse of personal data. 

Transparency and accountability 🙈

Transparency limitations associated with the Microsoft Celeb (MS-Celeb-1M) dataset include: 

• Unclear consent processes for individuals included in the dataset

• Lack of transparency about data collection methods and sources

• Limited information on how "celebrity" status was determined

• Insufficient disclosure of potential biases in the dataset

• Inadequate documentation of data quality and accuracy

• Lack of clear policies on data usage, sharing, and restrictions

• Limited transparency on the dataset's impact on privacy and civil liberties

• Insufficient information on data retention and deletion processes

• Unclear mechanisms for individuals to request removal from the dataset

• Limited public reporting on how the dataset has been used by researchers and companies.

Research, advocacy 🧮

• Peng K., Mathur A., Narayanan A. (2021). Mitigating Dataset Harms Requires Stewardship: Lessons from 1000 Papers

• Harvey, A., LaPlace, J. (2019). Exposing.ai

Investigations, assessments, audits 🧐

• Murgia M., Financial Times (2019). Who’s using your face? The ugly truth about facial recognition