Submission

Civilian protection through technology

Sentry increases access to and the accuracy of warnings on incoming airstrikes using a low tech approach that leverages big data.

Saeed Uri

Deputy Chief of Party - Programs

suri@srprogram.com

Middle East

USAID/OTI Syria Regional Program II

Peace, Stability and Transition

It is estimated that more than 450,000 people have been killed and at least 1.9 million wounded in Syria
between 2011 and 2015. By 2016, 61% of civilian deaths in Syria were caused by warplane shelling
(airstrikes), compared to only 10% in 2012. In 2016 alone, 87% of total civilian deaths in one province from January through September of this year attributed to warplane shelling. Included in these warplane attacks are indiscriminate barrel bombs and munitions banned by international law. There was no dependable way to warn civilians of an imminent attack. While civilian spotter networks (“observatories” made up of individual “observers”) watch the skies day and night, they had no way to rapidly or dependably communicate warnings to affected populations. Warnings were unpredictably transmitted via walkie-talkie repeater networks, word-of-mouth, social media, and air raid sirens. There was no consistent, dependable way to rapidly spread warnings to civilians in advance of attacks.

While the situation in Syria is difficult, the partner was able to develop a creative solution that uses low-technology and big data to save civilian lives.

Sentry

Syria

Machine Learning

August 2016 - ongoing

To address this problem, Hala, in coordination with Syria Civil Defense (SCD) and other key members of
Syrian civil society, deployed Sentry. Sentry securely aggregates validated observations from proven
civilian observatories, processes them through a predictive algorithm, then rapidly distributes accurate,
relevant warnings to as many civilians as possible, as fast as possible. Sentry was officially launched on
August 2nd, 2016 and currently generates warnings in Idlib province. The system is based on a mobile phone application that observers use to input their observations, the observations are then sent to a central server and analyzed by a monitor. While the data inputted includes the type of aircraft and heading, Sentry compares that data with after action data collected from previous airstrikes using a predictive algorithm. A prediction, that includes location and timing, is then shared on social and traditional media outlets.

The system has increased access to and accuracy of warnings provided to civilians on incoming aircraft. According to system data, on average Sentry provided nearly 8 minutes of warning to civilians prior to
an attack. This distribution of minutes of warning received is shown below. Of warnings sent, 82%
provided 5 or more minutes of warning time prior to potential attacks. Due to data availability, only
warnings sent between September 11-30, with a potential target zone, are included in this calculation.

Additional, the system has been used for ceasefire monitoring as part of international efforts to end the conflict in Syria. For example, during the September 2016 Cessation of Hostilities (CoH), Sentry continued to track warplane activity and accumulated data for reporting violations. Sentry data was shared with select governments, NGOs, and international media outlets to shed light on the fragility of the CoH.

The team is now working to connect the system to sirens that will be installed in population centers. This cost money and requires additional effort. This will help further increase access to the data.

A similar approach can be used to support civilian protection in other circumstances.

4a_Final Report.pdf