Cybersecurity research examines how intelligent systems change identity, attack surfaces, software dependencies, infrastructure exposure and the operating model for institutional resilience.
Security must extend beyond networks and endpoints to the identities, tools, models, data paths, permissions and automated actions surrounding enterprise AI.
Security boundaries are being redrawn
AI systems increasingly connect to sensitive data, applications, infrastructure and business processes. Agentic systems can select tools and initiate actions, which makes machine identity, scoped authority, tool permissions, observability and containment first-order cybersecurity controls.
Identity & authority
Persistent machine identity, credential boundaries, least privilege, delegated authority, revocation and auditable activity.
AI attack surfaces
Prompt injection, data leakage, malicious tool use, model supply-chain exposure, retrieval manipulation and unsafe automation.
Infrastructure security
Cloud, compute, model endpoints, data stores, secrets, network paths and operational dependencies that support intelligent systems.
Software supply chain
Libraries, models, agents, plugins, APIs, data providers and third-party services that expand institutional dependency and risk.
Detection & response
Telemetry, anomaly detection, incident classification, containment, rollback, recovery and evidence preservation.
Critical systems
Cyber resilience for infrastructure and high-consequence environments where AI failure can produce operational or safety impact.
Research focus
The program studies how cybersecurity architecture must adapt as AI becomes embedded in enterprise workflows. Research connects technical controls to institutional operating questions: who or what is authorized to act, which resources can be reached, how actions are observed, how exceptions are handled and how systems are contained when behavior departs from expected boundaries.
Institutional objective
The objective is not zero risk. It is a control architecture in which exposure is understood, authority is bounded, activity is observable, critical dependencies are resilient and incident response can operate at machine speed when necessary.