The Protection Assistant for Wildlife Security (PAWS) is an artificial intelligence (AI) software for wildlife protection. PAWS helps conservation managers plan more informed patrol strategies to combat wildlife poaching, illegal logging, and illegal fishing. Taking in past poaching records and data about the geography of the protected area, PAWS uses machine learning to predict poachers’ behavior. After learning a model of poaching behavior, PAWS produces poaching risk maps and suggests patrol routes for rangers.

This AI-based anti-poaching tool has been developed in collaboration with conservation specialists at the Uganda Wildlife Authority, Wildlife Conservation Society, and the World Wide Fund for Nature. In December 2018, we began conducting field tests in Srepok Wildlife Sanctuary (SWS) in Cambodia, an area identified as the most suitable site for tiger reintroduction in Southeast Asia. During the first month of field tests, rangers conducted patrols based on regions we suggested, during which they detected and removed 1,000 snares — more than double before the deployment of the AI software. Rangers also confiscated 42 chainsaws, 24 motorbikes, and one truck. Additionally, PAWS is being integrated with SMART conservation software, a monitoring and reporting tool used in over 800 protected areas worldwide

Motivation: the snaring crisis

Poaching of endangered species is reaching critical levels as the populations of these species plummet to unsustainable numbers. The global tiger population, for example, has dropped over 95% from the start of the 1900s and has resulted in three out of nine species extinctions. Poaching is often driven by demand for animal products including ivory, rhino horn, and leopard skin, which fuel a black market worth USD 20 billion per year. 

To counter poaching, governments have set up protected wildlife reserves and conservation agencies tasked with defending these large reserves. Because of the size of the reserves and the common lack of law enforcement resources, rangers are at a significant disadvantage when it comes to deterring poachers and removing snares. Agencies use patrolling as a primary method of securing the park. Due to their limited resources, however, patrol managers must carefully create patrols that account for many different variables. PAWS helps rangers locate high-risk areas and plan patrols subject to complex constraints. 

Global deployment of PAWS

SMART is a software used for protected area management by over 800 national parks around the globe. The SMART partnership is comprised of nine leading conservation NGOs including the World Wildlife Fund and Wildlife Conservation Society. In collaboration with SMART and Microsoft AI for Earth, we are making PAWS available to parks around the world to predict and plan against snaring, illegal logging, and illegal fishing. Read about how we're integrating PAWS into SMART to deploy our predictive models globally. To help under-resourced parks, we have also worked to integrate remote sensing imagery to augment their data and improve predictions. 

Multi-armed bandits to balance exploration and exploitation (AAAI'21) [paper]

Our paper Dual-Mandate Patrols: Multi-Armed Bandits for Green Security, which was selected as a Best Paper Runner Up at AAAI 2021, focuses on helping rangers choose where to patrol in a protected area, balancing exploration of infrequently visited regions and exploitation of known hotspots. Our proposed LIZARD algorithm bridges the gap between combinatorial and Lipschitz bandits, presenting an approach that achieves theoretical no-regret and performs well in experiments on real-world poaching data.

Predicting poaching, planning patrols, and evaluating in the field (ICDE'20) [paper]

Our ICDE 2020 paper Stay Ahead of Poachers: Illegal Wildlife Poaching Prediction and Patrol Planning Under Uncertainty with Field Test Evaluations addresses uncertainty in poaching predictions to help design patrol plans that are robust to uncertainty. We evaluate our predictions in months-long field tests in Murchison Falls National Park in Uganda and Srepok Wildlife Sanctuary in Cambodia. 

Adversary models account for imperfect crime data: Forecasting and planning against real-world poachers. (AAMAS'18) [paper]

In this innovative application paper, we introduce iWare-E a novel imperfect-observation aWare Ensemble (iWare-E) technique, which is designed to handle the uncertainty in crime information efficiently. This approach leads to superior accuracy and efficiency for adversary behavior prediction compared to the previous state-of-the-art. We also demonstrate the country-wide efficiency of the models and are the first to evaluate our adversary behavioral model across different protected areas in Uganda. Lastly, we provide a scalable planning algorithm to design fine-grained patrol routes for the rangers, which achieves up to 150% improvement in number of predicted attacks detected.

Security games to prevent poaching and illegal fishing (IJCAI'15) [paper]

When Security Games Go Green: Designing Defender Strategies to Prevent Poaching and Illegal Fishing was selected as a Outstanding Paper Award Winner in IJCAI-15 Computational Sustainability Track at IJCAI 2015. This paper introduces Green Security Games (GSGs), a novel game model for green security domains with a generalized Stackelberg assumption; provides algorithms to plan effective sequential defender strategies — such planning was absent in previous work; proposes a novel approach to learn adversary models that further improves defender performance; and (provides detailed experimental analysis of proposed approaches.

Introducing PAWS (AAMAS'14) [paper]

Adaptive resource allocation for wildlife protection against illegal poachers introduces the Protection Assistant for Wildlife Security (PAWS) as a joint deployment effort done with researchers at Uganda’s Queen Elizabeth National Park (QENP) with the goal of improving wildlife ranger patrols. While previous works have deployed applications with a game-theoretic approach (specifically Stackelberg Games) for counter-terrorism, wildlife crime is an important domain that promotes a wide range of new deployments. Additionally, this domain presents new research challenges and opportunities related to learning behavioral models from collected poaching data. To address these challenges, we contribute a behavioral model extension that captures the heterogeneity of poachers’ decision making processes. Second, we provide a novel framework, PAWS-Learn, that incrementally improves the behavioral model of the poacher population with more data. Third, we develop a new algorithm, PAWS-Adapt, that adaptively improves the resource allocation strategy against the learned model of poachers. Fourth, we demonstrate PAWS’s potential effectiveness when applied to patrols in QENP, where PAWS will be deployed.