Artificial Intelligence for Security Planning: A Game-Theoretic Approach
Implementing Organization
Indian Institute Of Technology Bombay
Principal Investigator
Prof. Swaprava Nath
Indian Institute Of Technology Bombay
swaprava@gmail.com
CO-Principal Investigator
Dr. Sujoy Kumar Bhore
Indian Institute Of Technology Bombay, Iit Po Powai,Maharashtra,Mumbai-400076
Project Overview
Ensuring the safety of critical infrastructure and human assets is a vital responsibility for security agencies worldwide. In India, this challenge spans diverse scenarios such as border infiltration, urban counter-terrorism, anti-smuggling operations, and women’s safety. A key objective across these domains is the efficient deployment of limited security resources to minimize risk. Conventional methods rely on static or incident-informed patrol planning, which are often insufficient against strategic adversaries who adapt by observing predictable patterns. This proposal seeks to design AI-powered security solutions using Stackelberg Security Games (SSGs), where the defender (security agency) commits to a randomized patrol plan and the attacker optimally responds after observing it. The goal is to build SSG-based models grounded in geospatial and operational data, incorporating threat assessments derived from historical incidents and expert judgment. These models will inform optimization techniques, including mixed-integer programming and online algorithms, for both static and dynamic environments, enabling real-time, adaptive threat mitigation strategies. While the proposal contributes to the theoretical development of security games, its central aim is to understand India-specific security settings and develop usable, generalizable AI systems for them. The envisioned system can be applied to a wide range of Indian contexts. For instance, dynamic border patrol routes can be created using terrain and historical infiltration data. In urban policing, beat scheduling and checkpoint allocation can respond to real-time crime patterns. In high-risk zones, convoy routing for anti-insurgency missions can be optimized for safety and unpredictability. This approach adds strategic randomness to patrol plans, making it harder for adversaries to exploit predictable behavior. By incorporating these AI-based tools into the workflows of agencies such as the Indian Army, BSF, and state police, the project aims to deliver deployable systems that support adaptive and defensible resource planning. Ultimately, it lays the groundwork for building intelligent, adversary-aware, and data-driven patrol systems tailored to India’s complex and evolving security landscape.