Its main challenge is to develop methodology capable of breaking through the existing software and hardware limitations.
The European Research Council (ERC) has announced the ERC Consolidator Grants recipients, among whom is the Barcelona Supercomputing Center’s (BSC) Sr. Researcher, Antonio J. Peña, leader of the Accelerators and Communications for HPC team. Peña will be the PI of the awarded ERC proposal HomE (Enabling Homomorphic Encryption of Deep Neural Network Models and Datasets in Production Environments), that aims at enabling the feasibility of privacy-preserving machine learning in untrusted environments such as cloud services. Until now, homomorphically-encrypted techniques (those that enable computations directly on encrypted data) have posed an unbearable overhead. HomE will put together HPC software optimization techniques and novel hardware designs, on top of Persistent Memory technology, to enable production-sized scenarios of homomorphically-encrypted deep learning.
Antonio J. Peña says: “This ERC grant will bring deep-learning users the possibility of leveraging their jobs with confidence in a cloud provider, even when data protection policies are in place (e.g., GDPR). I imagine a hospital being able to offload advanced cancer detection processes based on deep neural networks without concerns about the privacy of the medical records. It is going to be simply amazing."
HomE Challenge: developing methodology capable of breaking through the existing software and hardware limitations
Deep learning (DL) is widely used to solve classification problems previously unchallenged, such as face recognition, and presents clear use cases for privacy requirements. Homomorphic encryption (HE) enables operations upon encrypted data, at the expense of vast data size increase. RAM sizes currently limit the use of HE on DL to severely reduced use cases. Recently emerged persistent memory technology (PMEM) offers larger-than-ever RAM spaces, but its performance is far from that of customary DRAM technologies.
HomE project aims at sparking a new class of system architectures for encrypted DL workloads, by eliminating or dramatically reducing data movements across memory/storage hierarchies and network, supported by PMEM technology, overcoming its current severe performance limitations.
HomE intends to be a first-time enabler for the encrypted execution of large models that do not fit in DRAM footprints to execute local to accelerators, hundreds of DL models to run simultaneously, and large datasets to be run at high resolution and accuracy.
Targeting these ground-breaking goals, HomE enters into unexplored field resulting from the innovative convergence of several disciplines, where wide-ranging research is required in order to assess current and future feasibility. Its main challenge is to develop methodology capable of breaking through the existing software and hardware limitations. HomE proposes a holistic approach yielding highly impactful outcomes that include novel comprehensive performance characterisation, innovative optimisations upon current technology, and pioneering hardware proposals. HomE can spawn a paradigm shift that will revolutionise the convergence of the machine learning and cryptography disciplines, filling a gap of knowledge and opening new horizons such as DL training on HE, currently too demanding even for DRAM. HomE, based on solid evidence, will unveil the great unknown of whether PMEM is a practical enabler for encrypted DL workloads.
About Antonio J. Peña
Antonio J. Peña is a Sr. Researcher at the Barcelona Supercomputing Center, where he leads the Accelerators and Communications for HPC team. He is a prospective Ramón y Cajal Fellow and former Marie Sklodowska-Curie Individual Fellow. Among others, he is a recipient of the 2017 IEEE TCHPC Award for Excellence for Early Career Researchers in HPC, and ACM/IEEE Sr. Member. He is involved in the organization and steering committees of several conferences and workshops such as SC or IEEE Cluster. His research interests in the area of runtime systems and programming models for high-performance computing include resource heterogeneity and communications.