Around 50 people attended the FOT-NET Data workshop on Data Anonymization in Gothenburg, Sweden, on 31st August and 1st September 2016. The two-day workshop was an opportunity to gain insights into different ways of anonymization of FOT/NDS video data. These data which provide essential information on driver behaviour during different driving scenarios require a high level of data protection. The workshop covered multiple presentations by different stakeholders and discussions on various relevant topics including the role of data anonymization in vehicle automation.
During the first day of the workshop the importance of real world data for safety systems development from OEM point of view was discussed. Knowledge on how vehicles and people interact is essential when it comes to developing automated vehicles.
The main issues with anonymization and feature extraction were also discussed. As the amount of collected data is increasing fast, privacy and security principles must follow at the same pace. Protecting data privacy includes data protection, anonymization and feature extraction while the importance of computer resources for large-scale anonymization and feature extraction is not to be neglected.
The US perspective on anonymization and the operationalizing privacy was presented by Hannah Rakoff of the US DOT who highlighted advancing automation, connected vehicles and smart cities as the strategic priorities. It is important to work together in this field to make progress. The US is especially interested in methods to anonymise real-time trajectory data. Devdatt Dubhashi, Chalmers, pointed out the general concern about privacy in computer science and presented different privacy for data mining and querying.
On the second day of the workshop John Lee of the University of Wisconsin gave a presentation on feature extraction and how to assure you can delete data without information loss.
Ways of using automated video annotation for feature extraction were also discussed. Jan Erik Solem gave a presentation on automated labelling and recognition using the example of his Mapillary project which tries to solve scalability in the image collection and in the data extraction.
The second day concluded with the discussion on the core question: “What is anonymous enough?” which also highlighted the importance of ethics research in addition to the legal aspects.
You can download all the presentations in the library.