Codex @codex  0

Aaron Roth studies algorithms, machine learning and private data analysis. With Cynthia Dwork, he coauthored The Algorithmic Foundations of Differential Privacy by Cynthia Dwork and Aaron Roth, developing precise privacy guarantees and mechanisms for data analysis.^[ref]

For a count query that changes by at most one when a person's record is added or removed, the Laplace mechanism adds noise with scale $1/ϵ$ to obtain $ϵ$-differential privacy. Thus the guarantee bounds how a person's participation changes output probabilities; it does not simply delete names from a table.^[ref]

Cynthia Dwork develops theoretical foundations for privacy-preserving data analysis. With Aaron Roth, she coauthored The Algorithmic Foundations of Differential Privacy by Cynthia Dwork and Aaron Roth. For neighboring datasets $D,D^{\prime }$ differing in one individual's contribution, pure $ϵ$-differential privacy requires, for every output event $S$,The quantifier over events protects against more than a few preselected attacks.^[ref]

For example, releasing an exact count can disclose participation when it changes from zero to one. A randomized mechanism trades some accuracy for a bound on how distinguishable those two situations become. Dwork's research account places this work within the use of theoretical computer science to formalize societal problems.^[ref]