LATIDIA · Ciberseguridad

Inversión del modelo consciente del hardware en redes neuronales con picos de memoria

arXiv:2610.39688v1 Anuncio Tipo: nuevo Resumen: Las redes neuronales de picos (SNN) aceleradas por hardware implementadas en matrices de barras cruzadas memristivas ofrecen ganancias significativas en eficiencia energética, mientras que la variabilidad del dispositivo y la ana

La noticia

arXiv:2610.39688v1 Announce Type: new Abstract: Hardware-accelerated spiking neural networks (SNNs) implemented on memristive crossbar arrays offer significant gains in energy efficiency, while device variability and analog noise are sometimes assumed to hinder information extraction. However, whether these hardware effects improve, preserve, or degrade privacy after deployment remains largely unexplored. Therefore, we ask whether model-inversion leakage characterized on an SNN remains representative after the same model is mapped to a non-ideal memristive substrate, and how hardware non-idealities and reliability faults affecting synapses, neuron thresholds, and spike processing reshape that leakage. To study this problem, we introduce Spiking Domain Model Inversion (SDMI), an inversion method that combines Expectation over Transformation (EOT) with