THE CENTRE FOR HUM ANITARIAN DATA

GUIDANCE NOTE SERIES

DATA RESPONSIBILIT Y IN HUMANITARIAN AC TION

NOTE #8: RESPONSIBLE APPROACHES TO DATA

SHARING

KE Y TAKE AWAYS:
• Open sharing of timely and accurate data is essential to effective and efficient humanitarian
response. How humanitarian organizations approach data sharing directly relates to trust and
cooperation in the sector.

• As the humanitarian data ecosystem grows, the opportunities and risks of sharing data become
clearer, prompting organizations to explore more limited approaches to data sharing.

• Humanitarian organizations widely recognize the sensitivity of personal data: its exposure has a
high likelihood of causing harm. The majority of non-personal data is safe to share openly, but
non-personal data can also be sensitive and should be handled with caution.

• Humanitarian organizations should take into account four factors when deciding whether to
share non-personal data: (i) utility; (ii) sensitivity; (iii) human and technical capacity; and (iv)
governance.

• Humanitarian organizations should identify and compare all available approaches for data
sharing, considering the most open approach first and working down to more limited approaches
as necessary.

INTRODUC TION
The use and exchange of data have become core functions of humanitarian organizations. Staff regularly
need to decide whether and how to share their organization’s data, even if their role is not primarily
focused on data or information management. Beyond individual organizations, the interest in the
sharing and use of data generated in humanitarian action has also grown. In response to this interest, the
humanitarian sector has seen a surge in data generation and sharing in recent years.

“Accurate data is the lifeblood of good policy and decision-making.

Obtaining it, and sharing it across hundreds of organizations, in the middle
of a humanitarian emergency, is complicated and time-consuming — but it is
absolutely crucial. ”
- United Nations Secretary-General António Guterres at the opening of the OCHA Centre for
Humanitarian Data in The Hague in December 2017

Open sharing of timely and accurate data is essential to effective and efficient humanitarian response and
should remain a key objective for the sector. For example, the COVID-19 epidemiological data1 compiled
and shared daily by the Johns Hopkins University Center for Systems Science and Engineering has been
1
Access the Novel Coronavirus (COVID-19) Cases Data on HDX.

THE CENTRE FOR HUMANITARIAN DATA DECEMBER 2020 1

integrated in a number of dashboards and situation reports for decision-makers across the humanitarian
sector since the outset of the pandemic. By December 2020, this dataset had been downloaded more than
320,000 times from the Humanitarian Data Exchange (HDX) since it was first published on the platform in
January that year. The utility of open data to humanitarian practitioners is further underlined by the fact
that usage of HDX in countries with a Humanitarian Response Plan (HRP) in place has grown much faster
than usage in other locations.2

How humanitarian organizations approach data sharing directly relates to trust and cooperation in the
sector. Maintaining trust within the data ecosystem is critical to the sustainability of data sharing and
relates to issues such as the quality of the data, the level to which the data will be secured after sharing
and the responsible use of data by the recipient. Because data in the humanitarian sector often relates to
the most at-risk populations, managing and sharing it warrants caution.

Many humanitarian organizations have developed or updated their guidance, governance and practices
to support different aspects of data responsibility: the safe, ethical and effective management of data. The
sector has also seen an increasing number of collaborative efforts to improve data responsibility beyond
individual organizations.3 Still, as the humanitarian system learns more about the risks associated with
data sharing, organizations face more complex challenges in sharing this data responsibly.4

This guidance note aims to support decision-making around the sharing of non-personal data in
humanitarian settings. It explains data sensitivity, provides common examples of sensitive non-personal
data, and explains an approach to information and data sensitivity classification in humanitarian settings.
It also offers a framework that organizations can use to weigh four factors that help determine whether
data can be shared and explains common approaches for doing so responsibly.

Options for Data Sharing on the Humanitarian Data Exchange

When the Humanitarian Data Exchange (HDX) was launched in 2014, it held close to 900 datasets, shared
by a handful of ‘early adopter’ organizations. By the end of 2020, that number had grown to over 18,000
datasets. Only approved organizations are able to share data on the platform. They can make data
available publicly to anyone who visits the site or privately to only the members of their organizations.

In 2017, the HDX team added another option for data sharing: HDX Connect. This feature enables
organizations to only publish the metadata, with the underlying data available upon request. If access is
granted, the data is shared bilaterally without passing through the HDX platform. For example, Ground
Truth Solutions use HDX Connect to provide access to COVID-19 Community Perceptions Data collected in
Iraq.

As part of its quality assurance process, the HDX team also runs a disclosure risk assessment on any
resource added to the platform that contains microdata. The HDX team does this because it may be
possible to re-identify individuals or expose confidential information even after direct identifiers have been
removed from microdata.5

Some organizations on HDX have become more oriented towards controlled access to their data, either
due to the sensitive nature of the data, increased pressure to track and report on how the data is used,
or resource constraints related to operational sustainability. HDX will always support different ways of
sharing data — however, open access remains the best option for the majority of data that is generated for
humanitarian response.

2
“From August 2016 through August 2020 (the period for which the data is available), growth in monthly users from HRP+ countries was 943%
compared to 566% across all countries.” From the HDX Case Study, September 2020.
3
These include, for example, the IASC Sub-Group on Data Responsibility in Humanitarian Action, the Protection Information Management
initiative, and the Responsible Data for Children initiative, among others.
4
For a better understanding of the challenge facing humanitarian organizations when sharing data specifically in protracted humanitarian crises,
see ALNAP, Data Collection, Analysis and Use in Protracted Humanitarian Crises, June 2020.
5
Learn more about the Centre’s risk mitigation process for microdata, ‘Statistical Disclosure Control’ or SDC in the Learning Path on the topic.

THE CENTRE FOR HUMANITARIAN DATA DECEMBER 2020 2

UNDER S TANDING DATA SENSITIVIT Y
Humanitarian organizations widely recognize the sensitivity of personal data6: its exposure has a high
likelihood of causing harm. This understanding does not yet widely exist for non-personal data, which
typically covers the following three categories in humanitarian settings:

1. Data about the context in which a response is taking place (e.g. legal frameworks, political, social
and economic conditions, infrastructure, etc.) and the humanitarian situation (e.g security incidents,
protection risks, drivers of the situation or crisis).
2. Data about the people affected by the situation and their needs, the threats and vulnerabilities they
face, and their capacities.
3. Data about humanitarian response actors and their activities (e.g. as reported in 3W/4W/5W).

The majority of this data is safe to share openly. However, non-personal data can also be sensitive.
Examples of sensitive non-personal data include data on groups experiencing gender-based violence or the
location of ethnic minorities in conflict settings. Such data is considered sensitive because it enables the
identification of groups of individuals by demographically defining factors, such as ethnicity, gender, age,
occupation, religion or location of origin. Non-personal data can also create risk in other ways, for example
by exposing the location of medical facilities in areas where they are prone to attack. As the awareness of
the risk associated with sharing such data continues to grow, some organizations are turning from a focus
on open data to more controlled sharing.

Many organizations have information and data sensitivity classifications (see figure 1 below) that define
which data falls into which category of sensitivity in order to facilitate responsible data management. These
classifications may also be developed as a collective exercise to help organizations align around what
constitutes sensitive data in their context and identify the appropriate disclosure or dissemination methods
for different data types depending on their sensitivity.

Information and Data Sensitivity Classification7

Information and Data

Sensitivity Definition
Sensitivity Classification

Low or No Information or data that, if disclosed or accessed Public

without proper authorization, are unlikely to
cause any harm or negative impacts to affected
people and/or humanitarian actors.

Moderate Information or data that, if disclosed or accessed Restricted

without proper authorization, are likely to cause
minor harm or negative impacts and/or be disad-
vantageous for affected people and/or humani-
tarian actors.

High Information or data that, if disclosed or accessed Confidential

without proper authorization, are likely to cause
serious harm or negative impacts to affected peo-
ple and/or humanitarian actors and/or damage to
a response.

Severe Information or data that, if disclosed or accessed Strictly Confidential

without proper authorization, are likely to cause
severe harm or negative impacts and/or damage
to affected people and/or humanitarian actors
and/or impede the conduct of the work of a
response.

Figure 1. Sample Information and Data Sensitivity Classification7

6
Personal data should not be shared openly, and management of personal data should always comply with national and regional data protection laws, or with
internal data protection policies in the case of organizations covered by privileges and immunities.
7
UNOCHA (2019), Working Draft Data Responsibility Guidelines.

THE CENTRE FOR HUMANITARIAN DATA DECEMBER 2020 3

Organizational policies and collective governance instruments such as Information Sharing Protocols
(ISPs) often include a sensitivity classification and should be the primary points of reference for
determining how to manage sensitive data. However, these documents tend to leave room for discretion
regarding whether and how to share. This means data sharing can be influenced by personal preferences
and skills and may vary across organizations. By taking a more consistent approach to data sharing and
ensuring adequate safeguards for sensitive data, organizations can build trust and contribute to more
efficient and effective humanitarian response.

FOUR FAC TOR S FOR DETERMINING WHETHER TO SHARE NON - PER SONAL DATA

There are four factors humanitarian organizations should take into account when deciding whether to
share non-personal data.

1. What is the utility of the data for other stakeholders?

The utility of data depends on the level of detail, the number of people or the geographical area
covered, its timeliness, and its relevance to analysis and decision-making in humanitarian response.
Conduct a Data Impact Assessment (DIA) to help determine the utility of specific data.8

2. How sensitive is the data?

The sensitivity of data is based on the risk associated with its exposure in a particular context.9 In
some response contexts, organizations, clusters/sectors, and system-wide coordination structures
also have established data and information sensitivity classifications (see above) that can inform this
determination. Conducting a DIA can also help determine the sensitivity of data. For survey results
and other forms of microdata, sensitivity is closely linked to the risk of re-identification, which can
be determined by applying a disclosure risk assessment.

3. What human and technical capacity do the organizations sharing and using the data have?
Both the organization sharing data and the organization(s) receiving and using the data should
have sufficient human and technical capacity for responsible data management. This includes staff
availability, data literacy, technical infrastructure and related resources. In environments with low
connectivity, bandwidth-heavy data sharing methods may not be appropriate. For contexts with
known security risks, data should typically be shared through more limited approaches.

4. Which governance instruments apply?

Common data governance instruments include ISPs, data sharing agreements for bilateral data
sharing and licenses or terms of use for public data sharing.11 These instruments should inform how
data is shared in a safe, ethical and effective manner. In some situations, governance will need to
be developed for the selected data sharing approach. Governance instruments can address a range
of topics and special provisions, but should always include the following elements: (a) the purpose
and scope of sharing; (b) any limitations to how data should be managed after sharing; (c) roles and
responsibilities throughout the sharing process; and (d) procedures for data incident management.12

8
See the Guidance Note on Data Impact Assessments.
9
For data management in the humanitarian sector, risk can be defined as the likelihood and impact of harm resulting from data management.
10
The World Economic Forum, together with Washington University Centre for Information Assurance and Cybersecurity, the Sustainable
Development Solutions Network TReNDS and the NYU GovLab, have begun building a repository of data sharing agreements to support the
professionalization of this practice through their Contracts for Data Collaboration (C4DC) project.
11
The licenses recommended for data sharing via HDX are listed here: https://data.humdata.org/about/license.
12
For more information about data incident management in humanitarian response, see our Guidance Note on Data Incident Management.

THE CENTRE FOR HUMANITARIAN DATA DECEMBER 2020 4

IDENTIF Y AND COMPARE APPROACHES

Organizations should determine the best approach to data sharing based on the four factors above. These
approaches range from open sharing to maximize the benefit of data, to more limited approaches such
as bilateral data sharing or only sharing data insights. The table below contains an overview of different
approaches to data sharing and offers examples of some commonly used tools and platforms.

Approaches, Tools and Platforms for Data Sharing in Humanitarian Response 13

Approaches to Open Access Limited Access Bilateral Sharing No Data Sharing

Sharing
Sharing data Limiting access to Bilateral sharing is Data that should
publicly is the most data still allows the most limited not be shared at all
open approach to select partners to way in which data can still offer value
sharing, allowing use the data as long can be shared, to partners, if they
unmediated access as they meet certain directly with one are allowed to query
to anyone. requirements. partner. the data remotely
or benefit from it
indirectly.

Common Tools Organization data HDX Private HDX Connect OPAL17

and Platforms platforms UNHCR Microdata Email16 Aircloak18
HDX Library14 Dropbox Homomorphic
Open Listservs IFRC GO15 encryption19
Cluster/Sector Sharing Multi-Party
Computation20
Closed mailing lists

In comparing these different approaches, always consider the most open approach first and work down
to more limited approaches as necessary. Different data types will require different ways of sharing. For
example, large datafiles will require specialized infrastructure and Application Programming Interfaces
(APIs) are suitable for data that is published in the same format on a regular basis. Because technologies
for data sharing continue to evolve, organizations should regularly revisit and compare available data
sharing approaches.

13
Not all tools and platforms in this overview have been vetted by the UN Secretariat. Always consult the relevant Information Technology advisors
before using a new tool.
14
UNHCR’s MicroData Library.
15
IFRC’s GO Platform.
16
Within humanitarian responses, one of the most common ways to share data is via email attachments. When sharing data via email, always take
the necessary security precautions. This way of sharing is responsible in some cases, but there are often more suitable ways to share data. For
information on how to encrypt email, see for example: https://www.cloudwards.net/how-to-encrypt-your-emails/.
17
The Open Algorithms Project.
18
Aircloak Insights.
19
To learn more about homomorphic encryption as a way of sharing the value of sensitive data, see here: htps://www.microsoft.com/en-us/research/
project/homomorphic-encryption/ and here: https://www.wired.com/story/google-private-join-compute-database-encryption/.
20
To learn more about multi-party computation, see here: https://www.tno.nl/en/focus-areas/information-communication-technology/roadmaps/
data-sharing/secure-multi-party-computation/.

THE CENTRE FOR HUMANITARIAN DATA DECEMBER 2020 5

 Unlocking Value from Data Without Sharing It: The Querying Approach

A relatively new approach to utilizing data without transferring the data itself is ‘querying’. Querying
allows third parties to formulate specific questions to be asked of the data without accessing it directly.
The resulting insights can then be checked for sensitivity and any other issues by the holder of the data.
This approach avoids transfers of data which can cause legal and ethical concerns, while still allowing for
valuable insights to be used for public good.

In implementing a querying approach, it is critical to establish governance in the form of instructions and
boundaries regarding the queries that may be sent, in order to prevent retrieval of sensitive information by
posing a combination of questions.21 Vetting users as well as their questions should always be a key step in
the process around this type of approach.

Commercial solutions to set up querying approaches include Aircloak Insights, which acts as a ‘proxy
between analysts and the sensitive data they need to work with.’ Another querying tool is the Open
Algorithms (OPAL) platform. This tool was specifically developed for the humanitarian and development
sectors and is currently being piloted in Colombia.

Data Sharing Methods at the Joint IDP Profiling Service

In 2019, the Joint IDP Profiling Service (JIPS) was awarded a grant by the UNHCR Innovation Fund to
research advanced data science methods for data anonymization. JIPS studied methods such as Multi-
Party Computation and Homomorphic Encryption and worked with technical experts at the Johns Hopkins
University Applied Physics Laboratory, Flowminder and the Government of Colombia National Statistics
Office.

In close collaboration with Flowminder and building on their Flowkit, JIPS developed a prototype
querying approach to enable humanitarian and development actors to safely access and query sensitive
individual-level data without needing to share it. The team developed a technical workflow to demonstrate
the viability of this approach with one single data provider and mapped the problems and limitations in
case of multiple data providers.

Organizations are encouraged to share their experience in promoting responsible data sharing with the
Centre for Humanitarian Data via [email protected].

COLLABORATORS: JOINT IDP PROFILING SERVICE (JIPS).

The Centre for Humanitarian Data ('the Center'), together with key partners, is publishing a series of eight
guidance notes on Data Responsibility in Humanitarian Action over the course of 2019 and 2020. The Guidance
Note series follows the publication of the working draft OCHA Data Responsibility Guidelines in March 2019.
Through the series, the Centre aims to provide additional guidance on specific issues, processes and tools
for data responsibility in practice. This series is made possible with the generous support of the Directorate-
General for European Civil Protection and Humanitarian Aid Operations (DG ECHO).

This document covers humanitarian aid activities implemented with the financial assistance of the
European Union. The views expressed herein should not be taken, in any way, to reflect the official opinion
of the European Union, and the European Commission is not responsible for any use that may be made of
the information it contains.

This project is co-funded

by the European Union

21
For an explanation of this risk, see for example: https://www.usenix.org/conference/usenixsecurity19/presentation/gadotti.

THE CENTRE FOR HUMANITARIAN DATA DECEMBER 2020 6