Laser Fault Injection in Memristor-Based Accelerators for AI/ML and Neuromorphic Computing

2025-10-15

AI Summary

Memristor crossbar arrays (MCAs), while offering high density and analog in-memory computing advantages, introduce novel fault-injection attack surfaces due to their non-volatile physical characteristics. Method: This work presents the first systematic study of laser-based fault injection targeting analog-domain computational integrity in MCAs. We propose a fine-grained, memristor-specific laser attack model, integrating HSPICE simulations (45 nm CMOS) with laser-induced photocurrent mechanisms to enable analog-domain differential fault analysis (DFA). Contribution/Results: Our approach achieves highly accurate internal weight inference (99.7% accuracy), end-to-end model cloning, and targeted weight manipulation—inducing an average 143% weight increase—thereby compromising reliability in AI/ML and neuromorphic computing. This work uncovers a previously unrecognized physical-layer security threat to analog in-memory computing hardware, provides the first experimental validation of DFA efficacy in the analog domain, and delivers critical insights for secure neuromorphic chip design.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Adversarial Attacks & Robustness
• Cognitive Modeling & Cognitive Systems: Analogy

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Memristive crossbar arrays (MCA) are emerging as efficient building blocks for in-memory computing and neuromorphic hardware due to their high density and parallel analog matrix-vector multiplication capabilities. However, the physical properties of their nonvolatile memory elements introduce new attack surfaces, particularly under fault injection scenarios. This work explores Laser Fault Injection as a means of inducing analog perturbations in MCA-based architectures. We present a detailed threat model in which adversaries target memristive cells to subtly alter their physical properties or outputs using laser beams. Through HSPICE simulations of a large MCA on 45 nm CMOS tech. node, we show how laser-induced photocurrent manifests in output current distributions, enabling differential fault analysis to infer internal weights with up to 99.7% accuracy, replicate the model, and compromise computational integrity through targeted weight alterations by approximately 143%.

Problem

• Explores laser fault injection vulnerabilities in memristor-based AI accelerators
• Analyzes how laser-induced faults compromise computational integrity in crossbar arrays
• Demonstrates differential fault analysis to extract internal weights with high accuracy

Innovation

• Laser fault injection targets memristive crossbar arrays
• Photocurrent perturbations enable differential fault analysis
• Simulations reveal 99.7% weight inference accuracy