This article was originally published by UNHCR Innovation Service, on Medium. Click here to read the original article.
Part two of this series on JIPS’ efforts to securely access and share data to better serve refugees and IDPs takes a closer look at the project’s journey, challenges, and accomplishments.
By Amy Lynn Smith — Independent Writer + Strategist
For the UN Refugee Agency (UNHCR) and other humanitarian agencies, using data to understand the migration patterns and general activities of refugee and Internally Displaced People (IDP) populations is vital to anticipating their needs. As explained in part one of this series, data that’s as current as possible is essential to providing the right services at the right time.
But even more important than timely information is keeping that data secure, especially for refugees and IDPs who have fled their homes due to persecution, conflict, or disaster. Refugees and IDPs not only deserve privacy, but data security can help ensure their safety. UNHCR and its Innovation Service have been working with the Joint Internal Displacement Profiling Service (JIPS), an inter-agency entity that promotes collaborative data processes and works with a variety of partners. Together, they have been working to solve the challenge of the lack of knowledge and trust related to methods of data anonymization and sharing, as well as a lack of adequate approaches designed specifically for humanitarian purposes. Equally important, they want to test whether anonymization alone is sufficient to protect the data privacy of people affected by crises.
Ultimately, accessing certain types of humanitarian data proved too slow and cumbersome, so the team focused on experimenting with telecom data, such as cell phone and text records, to further their efforts to provide even better assistance while ensuring data anonymity. This privacy is important not only for the safety of refugees and IDPs, but to the mobile network operators (MNOs) that provide that data — not to mention other organizations that may provide data in the future, which is the kind of broader analysis JIPS hopes to provide when brought on to a project in the future.
JIPS was awarded a grant from the Innovation Service’s Innovation Fund to research and test new ways to safely share this data between humanitarian organizations. JIPS put together a team of external experts: Alexandre Poltorak, an expert in software engineering and encryption; Togglecorp, which specializes in software engineering and data science expertise; and Flowminder, a nonprofit foundation that uses mobile operator, geospatial, and survey data to support humanitarian and development initiatives.
The project is driven by a number of key questions: On behalf of its partners, how can JIPS safely use and analyze microdata — such as information about individuals or households — that’s behind the owner’s firewall, whether it’s a mobile network operator or government? How can the data produced through a profiling exercise to understand displaced people’s situations, needs and intentions be made available safely to all partners? How can everyone involved be certain the data search is anonymized when data is being analyzed or queried? And how can the results of the data search be safely exported?
“We ask these questions because this is indeed our daily bread and butter, says Wilhelmina Welsch, Acting Coordinator at JIPS. “We like to think that we have professionalized collaboration, which is a very soft and intangible thing. We are constantly looking for ways, tools, and processes to show that working together always outweighs the perceived benefits of not working together. And the technology out there that is currently available can support working together and show that it is the best, most secure, most cost-effective way of doing things.”
In phase one of the project, JIPS and its partners conducted research to document state-of-the-art data anonymization and privacy-preserving methods. After a desk review, in phase two the team developed a prototype workflow to allow those in the humanitarian and development field working with microdata to safely access and query sensitive individual-level data without needing to directly share the personal and non-personal details in the underlying data. The team then developed a workflow using Call Data Records (CDRs) demonstrating that it is possible to perform data analysis within the data provider’s infrastructure and only export anonymized and aggregated data.
Imagining improved solutions through trial and error
As anyone knows who has ever ventured into unknown territory — which, for JIPS and most humanitarian actors this project is to a large degree — there are often detours and even setbacks. Welsch says there were a number of times the team thought they were on the right track but then “hit the wall” with its expectations on exactly what might be possible. After all, any novel idea requires experimentation as part of the innovation process.
According to Welsch, the team repeatedly thought they had convinced partners to share their own training data, technically known as “dummy data,” or agree to participate in testing the prototype to show real proof of concept, but then did not get the green light. The team was also constrained by the administrative timelines of UNHCR’s Innovation Fund, which required specific milestones to adhere to the organization’s yearly financial cycle. What’s more, building trust takes time and can’t be rushed. Time also proved a constraint when bringing on partners to test the new technology and working with some of the most precious resources today: data.
“We had to change direction several times in the course of the project,” Welsch says. “At first, it was a bit of a haphazard sort of journey. But we knew there were others with systems in place that resembled what we wanted to achieve but were designed for a different purpose. Flowminder’s work was one of them.”
For example, the project team used FlowKit software, developed by Flowminder, which provides secure access, processing, and analysis of anonymized mobile data to present a simulation of raw telecom data to mobile network operators.
“One of the major initial stumbling blocks was the hope that there was going to be at least one agency that could provide the kind of data about displaced people that JIPS is interested in,” says Jonathan Gray, FlowKit Lead Engineer and Data Analyst at Flowminder. “But we were able to support them in generating dummy mobile operator data and demonstrating a secure, privacy-respecting approach to analysis in that domain. That showed them that we can actually get aggregate data from mobile operator data, and they were very keen to adapt to what was practical and possible.”
According to Rebeca Moreno Jimenez, an Innovation Officer (Data Scientist) with UNHCR’s Innovation Service, the need to be flexible is an element of innovation. Making the most out of the data of mobile operators, especially CDRs, and creating strong partnerships like the one with Flowminder — as well as taking the time necessary to build trust — and incorporating it into JIPS’ original project, is a perfect example of considering experimentation as a stepping stone, not a failure.
“Part of the journey is tweaking an idea, often multiple times, and then coming up with brilliant solutions to test, always with an end goal in mind,” she says. “Some might consider it a failure when the initial goal is not achieved, but when it happens it keeps the reality of the humanitarian sector in place. And to overcome that then shows great resilience on the team — and creates a good mindset for success in the future.”
Embracing innovation as a strategy to securely find and share new data sources
For Welsch, the innovation process, which is rarely, if ever, linear, has been eye-opening. “We were hoping to work with humanitarian data that is as close as possible to real-life situations,” she says. “Although there was enthusiasm about the possibility of new data-sharing methods, we ended up implementing the work with a different type of data. We were extremely ambitious but realized with humility that it is the best we can do now and that we need to move forward from here. Now that we have a prototype that’s ready to use, we hope the project entices others to try it with humanitarian data.”
Gray adds that the process of experimentation with JIPS and the rest of the team led to the future potential to tackle the wider solution, which is safe and effective information-sharing between multiple organizations. Together, the team will continue working on ways to access the kind of contextual data JIPS’ partners need the most.
“We’re now at the point where we’ve tested a couple of different things and have a suggested methodology of how this could work,” Welsch says. “There’s still more space to explore, and through collaborations like this we can continue to build trust and realize the potential to be even more effective and rigorous in our response to supporting the lives and situations of people who are displaced.”