Simulators
Efficient simulation of spiking neural networks that takes advantage of their sparse, event-driven computation.
AISx Edge Intelligence Lab
Research
Lab Research Agenda
Our research investigates neural network–based intelligence at the edge, covering foundational questions and use-inspired work motivated by security needs. It is organized around three areas: bio-inspired spiking neural networks and neuromorphic computing, edge deep learning for nuclear nonproliferation, and edge deep learning for cybersecurity.
01
Neuromorphic Computing at the Edge
Spiking neural networks that compute efficiently and adapt to changing conditions, from simulation and learning algorithms to deployment on edge hardware.
Learning Algorithms
Bio-inspired and evolutionary algorithms for learning in spiking neural networks.
Continual Adaptation
Reward-based spike-timing-dependent plasticity (STDP) and Hebbian learning mechanisms that enable ongoing adaptation to changing conditions.
Edge Deployment
Connecting algorithms to low-power neuromorphic and conventional hardware platforms.
02
Edge Intelligence for Cybersecurity
Local AI systems for cybersecurity analysis, digital forensics, and intrusion detection, alongside research on the security and reliability of edge AI itself.
Local Small Language Models for Cybersecurity
Small Language Models that can run locally and support cybersecurity needs including digital forensics, reverse engineering, and malware analysis.
File Carving
Local deep learning systems for more efficient file carving in forensic investigations and data recovery.
Neuromorphic Intrusion Detection
Neuromorphic systems for detecting intrusions in networks and Controller Area Network (CAN) communications.
Security of Edge AI Systems
Technologies to secure AI systems operating at the edge and improve their reliability under adversarial conditions.
03
Edge Intelligence for Nuclear Nonproliferation
Deep learning and neuromorphic methods for radiation detection, nuclear safeguards, remote sensing, and multimodal data fusion at the edge.
Near-Field Detection with Gamma Spectroscopy
Deep learning models for radioactive source search, radioisotope identification, and source localization using gamma spectroscopy.
Neuromorphic Computing in Safeguards
Combining radiation and camera data with neuromorphic computing to support nuclear safeguards.
Deep Learning in Remote Sensing
Deep learning models for anomaly detection at long standoff distances.
Multimodal Data Fusion at the Edge
Combining complementary sensor data at the edge to support nuclear nonproliferation analysis and detection.