Resource Library

Products

Threat Deterrence in Soft Urban Transport Targets with Network Science and Machine Learning

To deter adversaries from targeting soft transportation infrastructure network targets, such as rail transit systems, the NU-SDS laboratory has developed a network risk modeling and simulation capability that leverages the power of Artificial Intelligence (AI), risk science, and network science, and incorporates network structure, dynamics, and cascading effects to assess infrastructure network threat risk and deterrence and inform resource allocation decisions. This generalizable modeling and simulation capability, applied to the Boston urban rail network as a proof-of-concept, includes three connected computational modules: (1) machine learning with graphs – with graph convolutional networks (GCN) for target attractiveness prediction, (2) network failure simulation – with combinatorial methods to simulate cascading network failure sequences for what-if scenario analysis, and (3) network resource optimization – with nested defender-adversary optimization methods for risk-informed network resource allocation.

Threat Deterrence in Soft Urban Transport Targets

References for Layered Security and Architectural Design

This document contains citations for reference materials collected during the Architectural Design Research: Integrating Security in the Public Realm project research through the date listed in the document.  No endorsement or critique is made or implied by the presence or absence of any source on this list. Citations are grouped into libraries to support navigation and listed alphabetically within each heading, e.g., CPTED, Environmental and Criminal Psychology, School Security, Transportation Security, etc.. Items may be duplicated in multiple locations. Good faith efforts have been made to provide complete and correct citations, but they may contain inadvertent errors.

Project References

Assessing Agent and Environment Interactions for Multimodal Human Trajectory Prediction

Existing human trajectory datasets cannot capture interactions. They either lack Agent-to-Environment (A2E) interactions or exhibit limited Agent-to-Agent (A2A) interactions at a small scale in simple environments. We speculate that many self-centered pedestrians are prone to consciously or unconsciously avoiding or mitigating the influence of the environment and other pedestrians during navigation. In this work, we propose datasets that augment A2E and A2A interactions, which may benefit learning models by encoding more complex dynamics in trajectories. We propose a comprehensive trajectory prediction dataset, A-to-X, consisting of a representative set of trajectories that will enable better generalization under realistic, complex, or unsafe conditions and out-of-distribution with respect to current datasets.

HTP Benchmark Resources

Annual Reports

SENTRY Final Report Narrative

We’re proud to share our Final Report Narrative, which showcases the many accomplishments of the SENTRY program during it’s almost 4 years funded by DHS. This report showcases our progress and achievements through April 8, 2025, when funding lapsed. Our mission-critical research, focused on protecting soft targets and crowded places, will continue.

SENTRY Year 3 Annual Report Narrative

We’re excited to share our Year 3 Annual Report Narrative, which highlights the four key research areas, case studies, workforce and professional development initiatives, and other aspects of the SENTRY program. This report showcases our progress and achievements over the past year, underscoring our dedication to safeguarding soft targets and crowded places.

SENTRY Year 2 Annual Report Narrative

We are delighted to present our Year 2 Annual Report Narrative, an overview of the four major research areas, case studies, the workforce and professional development program, and other elements of the SENTRY program. This report reflects our growth and accomplishments over the last year and demonstrates our commitment to protecting soft targets and crowded places.