Catalogue Search | MBRL
Are you sure you want to remove the book from the shelf?
{{itemTitle}}
9
result(s) for
"Kim, Duhyeong"
Sort by:
Harnessing the potential of shared data in a secure, inclusive, and resilient manner via multi-key homomorphic encryption
In this manuscript, we develop a multi-party framework tailored for multiple data contributors seeking machine learning insights from combined data sources. Grounded in statistical learning principles, we introduce the Multi-Key Homomorphic Encryption Logistic Regression (MK-HELR) algorithm, designed to execute logistic regression on encrypted multi-party data. Given that models built on aggregated datasets often demonstrate superior generalization capabilities, our approach offers data contributors the collective strength of shared data while ensuring their original data remains private due to encryption. Apart from facilitating logistic regression on combined encrypted data from diverse sources, this algorithm creates a collaborative learning environment with dynamic membership. Notably, it can seamlessly incorporate new participants during the learning process, addressing the key limitation of prior methods that demanded a predetermined number of contributors to be set before the learning process begins. This flexibility is crucial in real-world scenarios, accommodating varying data contribution timelines and unanticipated fluctuations in participant numbers, due to additions and departures. Using the AI4I public predictive maintenance dataset, we demonstrate the MK-HELR algorithm, setting the stage for further research in secure, dynamic, and collaborative multi-party learning scenarios.
Journal Article
Privacy-preserving approximate GWAS computation based on homomorphic encryption
Background
One of three tasks in a secure genome analysis competition called iDASH 2018 was to develop a solution for privacy-preserving GWAS computation based on homomorphic encryption. The scenario is that a data holder encrypts a number of individual records, each of which consists of several phenotype and genotype data, and provide the encrypted data to an untrusted server. Then, the server performs a GWAS algorithm based on homomorphic encryption without the decryption key and outputs the result in encrypted state so that there is no information leakage on the sensitive data to the server.
Methods
We develop a privacy-preserving semi-parallel GWAS algorithm by applying an approximate homomorphic encryption scheme HEAAN. Fisher scoring and semi-parallel GWAS algorithms are modified to be efficiently computed over homomorphically encrypted data with several optimization methodologies; substitute matrix inversion by an adjoint matrix, avoid computing a superfluous matrix of super-large size, and transform the algorithm into an approximate version.
Results
Our modified semi-parallel GWAS algorithm based on homomorphic encryption which achieves 128-bit security takes 30–40 minutes for 245 samples containing 10,000–15,000 SNPs. Compared to the true
p
-value from the original semi-parallel GWAS algorithm, the
F
1
score of our
p
-value result is over 0.99.
Conclusions
Privacy-preserving semi-parallel GWAS computation can be efficiently done based on homomorphic encryption with sufficiently high accuracy compared to the semi-parallel GWAS computation in unencrypted state.
Journal Article
Essays on Monetary Economics
My dissertation, which consists of three papers, is devoted to studying the implications of conventional and unconventional monetary policies for inflation, asset prices, and welfare.The first paper examines the sustainability and effectiveness of negative nominal interest rates. I construct a model of multiple means of payment where the cost of holding paper currency—its storage and security costs—determines the effective rate of return on currency, which establishes the effective lower bound on nominal interest rates. I show that central banks can reduce the effective rate of return on currency, and thus the effective lower bound, by altering their policy on bank reserves. However, reducing the lower bound leads to welfare losses associated with individuals holding more currency. Moreover, sustaining a negative rate by reducing the lower bound has no stimulative effects. This occurs because this policy combination reduces both the rate of return on currency and interest rates on financial assets, leaving the relative interest rates between currency and financial assets unchanged.In the second paper, I develop a two-country model with financial frictions to study how a central bank’s unconventional asset purchases affect international asset prices and welfare. In the model, the key financial frictions are limited commitment, differential pledgeability of assets as collateral, and a scarcity of collateralizable assets. Due to the differential pledgeability of assets, financial intermediaries acquire different asset portfolios depending on their home country. I find that quantitative easing can reduce long-term bond yields and term premia internationally and depreciate the creditor country’s currency. Foreign exchange intervention always depreciates the local currency, but it can improve welfare globally if implemented by the creditor country.The third paper studies the implications of heterogeneous payment choices for monetary policy. I construct a model of money and credit where each consumer participates in a small-value or a large-value transaction depending on a preference shock. Financial intermediaries write deposit contracts for consumers to intermediate credit transactions. The preference shock is private information and is costly for intermediaries to observe. I find that, in equilibrium, financial intermediaries create state-contingent deposit contracts for consumers. However, private information and costly monitoring generate an incentive problem, so that the quantity of credit is constrained for consumers in large-value transactions. The effects of monetary policy on the allocation of means of payment vary depending on the size of transaction.
Dissertation
On the Scaled Inverse of \\((x^i-x^j)\\) modulo Cyclotomic Polynomial of the form \\(\\Phi_{p^s}(x)\\) or \\(\\Phi_{p^s q^t}(x)\\)
The scaled inverse of a nonzero element \\(a(x)\\in \\mathbb{Z}[x]/f(x)\\), where \\(f(x)\\) is an irreducible polynomial over \\(\\mathbb{Z}\\), is the element \\(b(x)\\in \\mathbb{Z}[x]/f(x)\\) such that \\(a(x)b(x)=c \\pmod{f(x)}\\) for the smallest possible positive integer scale \\(c\\). In this paper, we investigate the scaled inverse of \\((x^i-x^j)\\) modulo cyclotomic polynomial of the form \\(\\Phi_{p^s}(x)\\) or \\(\\Phi_{p^s q^t}(x)\\), where \\(p, q\\) are primes with \\(p
Paper
Ultra-Fast Homomorphic Encryption Models enable Secure Outsourcing of Genotype Imputation
Genotype imputation is a fundamental step in genomic data analysis such as GWAS, where missing variant genotypes are predicted using the existing genotypes of nearby \"tag\" variants. Imputation greatly decreases the genotyping cost and provides high-quality estimates of common variant genotypes. As population panels increase, e.g., the TOPMED Project, genotype imputation is becoming more accurate, but it requires high computational power. Although researchers can outsource genotype imputation, privacy concerns may prohibit genetic data sharing with an untrusted imputation service. To address this problem, we developed the first fully secure genotype imputation by utilizing ultra-fast homomorphic encryption (HE) techniques that can evaluate millions of imputation models in seconds. In HE-based methods, the genotype data is end-to-end encrypted, i.e., encrypted in transit, at rest, and, most importantly, in analysis, and can be decrypted only by the data owner. We compared secure imputation with three other state-of-the-art non-secure methods under different settings. We found that HE-based methods provide full genetic data security with comparable or slightly lower accuracy. In addition, HE-based methods have time and memory requirements that are comparable and even lower than the non-secure methods. We provide five different implementations and workflows that make use of three cutting-edge HE schemes (BFV, CKKS, TFHE) developed by the top contestants of the iDASH19 Genome Privacy Challenge. Our results provide strong evidence that HE-based methods can practically perform resource-intensive computations for high throughput genetic data analysis. In addition, the publicly available codebases provide a reference for the development of secure genomic data analysis methods. Competing Interest Statement The authors have declared no competing interest. Footnotes * Author name updated to fix a misspelling at the time of submission. The submitted PDF is already correct.
Paper
Myeloid compartment reprogramming through nanoparticle delivered resiquimod blocks paracrine growth support and activates phagocytosis to slow tumor progression in endogenous mouse medulloblastoma and diffuse midline glioma models
In pediatric brain tumors medulloblastoma (MB) and diffuse midline glioma (DMG), tumor-associated myeloid cells (TAMs) support malignant progression by secreting paracrine growth factors and suppressing local immune function. We studied the potential for reversing this cancer-supportive phenotype by stimulating TAM pathogen receptors using ResiPOx, a brain-permeant, polyoxazoline nanoparticle formulation of the TLR7/8 agonist resiquimod. ResiPOx showed blood-brain barrier penetration and anti-tumor efficacy, extending progression-free survival (PFS) in mice with MB and DMG. Integrated cellular and molecular analysis including scRNA-seq showed that ResiPOx expanded TAM populations and reprogrammed TAMs toward anti-tumoral states, blocking paracrine IGF1 signaling and inducing local cytokine signaling and phagocytosis of tumor cells. In rhesus macaques, systemic ResiPOx was well tolerated and induced brain transcriptional patterns that resembled ResiPOx responses in DMG and MB mouse models, indicating effects in non-human primates that highlight translational potential. Our data show that ResiPOx reshapes the brain tumor microenvironment to inhibit tumor growth. As a systemically administered, brain penetrant immunomodulator, ResiPOx is able to reach multifocal and unresectable brain tumors, including MB and DMG.
Journal Article
Nanoparticle-delivered resiquimod induces brain tumor regression in medulloblastoma and diffuse midline glioma models by interrupting paracrine growth support and activating myeloid immune signaling and phagocytosis
We studied the effect of stimulating innate immune function in tumor-associated myeloid cells (TAMs) in medulloblastoma (MB) and diffuse midline glioma (DMG), using a polyoxazoline nanoparticle formulation of the TLR7/8 agonist resiquimod (ResiPOx). Children with MB and DMG need novel therapeutic strategies to improve outcomes and reduce recurrence. We investigated the effect of systemically administered ResiPOx on TAMs in MB and DMG using endogenous MB and DMG models in immune-competent mice and identified multiple mechanisms of anti-tumor effect. We packaged resiquimod into polyoxazoline micelles to generate ResiPOx. We studied ResiPOx efficacy as a single agent or paired with radiation therapy (RT). We determined ResiPOx pharmacokinetics (PK) using tritium-labeled resiquimod and mass spectroscopy imaging (MSI). We determined ResiPOx pharmacodynamics (PD) using flow cytometry immunohistochemistry, bulk and single-cell RNA-seq and immunoblotting. We then studied ResiPOx safety and PD in a non-human primate model using rhesus macaques. ResiPOx formulation improved the blood-brain barrier penetration and anti-tumor efficacy of resiquimod. ResiPOx treatment extended progression-free survival (PFS) in mice with MB and DMG. In both tumor types, ResiPOx expanded TAM populations and reprogrammed TAMs toward anti-tumoral states, characterized by activation of IFNβ and extrinsic apoptosis pathway signaling, antigen presentation, and T cell activation signatures. In rhesus macaques, systemic ResiPOx administration was well tolerated and induced brain transcriptional responses that resembled ResiPOx responses in DMG and MB mouse models, indicating common effects across species from mice to non-human primates, and highlighting potential for similar effects in patients. ResiPOx is a brain-penetrant immunomodulatory therapeutic that reshapes the immune-privileged brain tumor microenvironment. Systemic administration activates myeloid-driven anti-tumoral immunity mediated by microglial and macrophage TAMs, and improves survival in preclinical models of DMG and MB.
Journal Article
Hospitality-VQA: Decision-Oriented Informativeness Evaluation for Vision-Language Models
Recent advances in Vision-Language Models (VLMs) have demonstrated impressive multimodal understanding in general domains. However, their applicability to decision-oriented domains such as hospitality remains largely unexplored. In this work, we investigate how well VLMs can perform visual question answering (VQA) about hotel and facility images that are central to consumer decision-making. While many existing VQA benchmarks focus on factual correctness, they rarely capture what information users actually find useful. To address this, we first introduce Informativeness as a formal framework to quantify how much hospitality-relevant information an image-question pair provides. Guided by this framework, we construct a new hospitality-specific VQA dataset that covers various facility types, where questions are specifically designed to reflect key user information needs. Using this benchmark, we conduct experiments with several state-of-the-art VLMs, revealing that VLMs are not intrinsically decision-aware-key visual signals remain underutilized, and reliable informativeness reasoning emerges only after modest domain-specific finetuning.
Paper
Advancing Small-Molecule Immunotherapy Through Polymeric Micelle Delivery
Small-molecule immunomodulators have become important components of modern immunotherapy by targeting immune checkpoints, cytokine signaling pathways, metabolic enzymes, and intracellular kinases. Despite pharmacological rationale, many of these agents underperform clinically due to unfavorable physicochemical properties, rapid systemic clearance, limited target accumulation, and dose-limiting toxicities, reflecting inadequate exposure control rather than a lack of target validity. Polymeric micelles, formed through the self-assembly of amphiphilic block copolymers, offer a versatile delivery platform to address these challenges by enhancing solubility, modulating pharmacokinetics, enabling stimuli-responsive release, and facilitating targeted or synchronized co-delivery. In this review, we classify representative small-molecule immunomodulators according to their immunological targets and examine the delivery constraints that shape their therapeutic performance. We then discuss design principles of polymeric micelle systems, including solubilization-driven formulations, microenvironment-responsive architectures, spatial targeting strategies, and co-delivery approaches that align cytotoxic and immunomodulatory mechanisms. Attention is given to the distinction between direct immunomodulators and cytotoxic agents that induce immunogenic cell death, highlighting how micelle-based delivery can enhance efficacy through improved exposure control. By integrating immunopharmacology with formulation science, this review outlines how polymeric micelles may advance the efficacy and safety of small-molecule immunomodulators and identifies key considerations for future translational development.
Journal Article