Search for: All records

In collaborative learning, multiple parties contribute their datasets to jointly deduce global machine learning models for numerous predictive tasks. Despite its efficacy, this learning paradigm fails to encompass critical application domains that involve highly sensitive data, such as healthcare and security analytics, where privacy risks limit entities to individually train models using only their own datasets. In this work, we target privacy-preserving collaborative hierarchical clustering. We introduce a formal security definition that aims to achieve balance between utility and privacy and present a two-party protocol that provably satisfies it. We then extend our protocol with: (i) an optimized version for single-linkage clustering, and (ii) scalable approximation variants. We implement all our schemes and experimentally evaluate their performance and accuracy on synthetic and real datasets, obtaining very encouraging results. For example, end-to-end execution of our secure approximate protocol for over 1M 10-dimensional data samples requires 35sec of computation and achieves 97.09% accuracy. 

In collaborative learning, multiple parties contribute their datasets to jointly deduce global machine learning models for numerous predictive tasks. Despite its efficacy, this learning paradigm fails to encompass critical application domains that involve highly sensitive data, such as healthcare and security analytics, where privacy risks limit entities to individually train models using only their own datasets. In this work, we target privacy-preserving collaborative hierarchical clustering. We introduce a {formal security definition} that aims to achieve balance between utility and privacy and present a two-party protocol that provably satisfies it. We then extend our protocol with: (i) an {optimized version for single-linkage clustering}, and (ii) {scalable approximation variants}. We implement all our schemes and experimentally evaluate their performance and accuracy on synthetic and real datasets, obtaining very encouraging results. For example, end-to-end execution of our secure approximate protocol for over 1M 10-dimensional data samples requires 35 sec of computation and achieves 97.09\% accuracy. 

Transparency logs allow users to audit a potentially malicious service, paving the way towards a more accountable Internet. For example, Certificate Transparency (CT) enables domain owners to audit Certificate Authorities (CAs) and detect impersonation attacks. Yet, to achieve their full potential, transparency logs must be bandwidth-efficient when queried by users. Specifically, everyone should be able to efficiently look up log entries by their key and efficiently verify that the log remains append-only. Unfortunately, without additional trust assumptions, current transparency logs cannot provide both small-sized lookup proofs and small-sized append-only proofs. In fact, one of the proofs always requires bandwidth linear in the size of the log, making it expensive for everyone to query the log. In this paper, we address this gap with a new primitive called an append-only authenticated dictionary (AAD). Our construction is the first to achieve (poly)logarithmic size for both proof types and helps reduce bandwidth consumption in transparency logs. This comes at the cost of increased append times and high memory usage, both of which remain to be improved to make practical deployment possible. 

Sensory neuron numbers and positions are precisely organized to accurately map environmental signals in the brain. This precision emerges from biochemical processes within and between cells that are inherently stochastic. We investigated impact of stochastic gene expression on pattern formation, focusing on senseless (sens), a key determinant of sensory fate in Drosophila. Perturbing microRNA regulation or genomic location of sens produced distinct noise signatures. Noise was greatly enhanced when both sens alleles were present in homologous loci such that each allele was regulated in trans by the other allele. This led to disordered patterning. In contrast, loss of microRNA repression of sens increased protein abundance but not sensory pattern disorder. This suggests that gene expression stochasticity is a critical feature that must be constrained during development to allow rapid yet accurate cell fate resolution.