This project examines the potential role of citizen-based monitoring and verification in a future, where digital data from from a variety of sources are abundant and widely available to non-governmental experts and independent analysts. Data authentication has always been a difficult aspect of open-source analysis, but as we enter an era where virtually any type of digital media can be generated in ways that can make them effectively indistinguishable from real data, the relevance of synthetic media in monitoring and verification deserves a careful and systematic analysis.
One part of this project seeks to systematically assess the long-term opportunities for citizen-based monitoring using the important example of satellite imagery in the context of nuclear monitoring and verification. Leveraging advanced machine-learning techniques to generate synthetic imagery of relevant sites, we can produce dedicated datasets under carefully controlled conditions. This imagery is then used to develop and examine concrete monitoring scenarios.
Another part of the project seeks to understand the risks and challenges, often enabled by the very same techniques; it emphasizes qualitative research methods and includes in-depth semi-structured interviews and a hands-on interactive tabletop exercise with focus groups and data users. In doing so, this part of the project also examines future challenges for citizen-based monitoring. We will place a particular emphasis on the possibility of image spoofing and fabricated data, examine broader ethical issues related to persistent earth observation, but also consider safeguards that could make citizen-based monitoring a viable and robust tool in support of peace & security.
Policy Relevance
This project offers a first systematic assessment of emerging monitoring and verification capabilities enabled by new satellite constellations and very short revisit times, using specially generated synthetic satellite imagery to do so. So far, satellite imagery has been considered "inherently trustworthy" because of the expense in creating it, but some experts have begun to warn about the potential threats of deepfakes to geospatial data and the possibility of satellite image spoofing for malicious purposes. One challenge for policymaking will be to avoid a situation, where anything could be "fake" and, by default, nothing may be considered authentic or credible. This project examines both technical strategies and possible policy interventions to avoid such an outcome.
Expected Research Results
There is a growing interest in the machine-learning and earth-observation communities to leverage synthetic imagery for research purposes. As part of this project, we will generate both curated datasets and, more importantly, make available the underlying framework so that other researchers can create additional datasets in support of their own work and assessments.
We are particularly interested in monitoring compliance with nuclear arms-control agreements, but similar opportunities are also emerging in several other contexts where international monitoring and verification are relevant. It has so far been difficult to systematically assess the true long-term potential and limits of citizen-based monitoring in this context. This project seeks to close this gap. It will also provide guidance and recommendations to ensure that the full potential of citizen-based monitoring can can be realized in the era of synthetic media and deepfakes. As researchers from Stanford University's Institute for Human-Centered Artificial Intelligence observed recently: "How we distinguish reality from the synthetic in our evolving world of thinking machines presents one of the most pressing questions of our time."
Please contact Alex Glaser for further information.