Home Page Contact Us Inst

Relevant Projects

Photo of Eitan Yaacobi
Associate Professor
Function-Correcting Codes

Motivated by applications in machine learning and archival storage, we introduce function-correcting codes (FCCs), a new class of codes to protect a function evaluation of the data against errors. We show that FCCs are equivalent to irregular-distance codes, i.e., codes that obey some given distance requirement between each pair of codewords. Using these connections, we study these codes and derive general upper and lower bounds on their optimal redundancy. Since these bounds depend on the specific function, we provide simplified, suboptimal bounds that are easier to evaluate.

Weakly Private Information Retrieval

Private information retrieval (PIR) protocols make it possible to retrieve a file from a database without disclosing any information about the identity of the file being retrieved. While existing protocols strictly impose that no information is leaked on the file’s identity, this project initiates the study of the tradeoffs that can be achieved by relaxing the requirement of perfect privacy. We propose to study this problem when the database is either replicated or is stored distributively over several servers, and when it is simply stored by a single server.

Weakly Private Information Retrieval

Private information retrieval (PIR) protocols make it possible to retrieve a file from a database without disclosing any information about the identity of the file being retrieved. While existing protocols strictly impose that no information is leaked on the file’s identity, this project initiates the study of the tradeoffs that can be achieved by relaxing the requirement of perfect privacy. We propose to study this problem when the database is either replicated or is stored distributively over several servers, and when it is simply stored by a single server.

Reconstruction Algorithms for DNA-Storage Systems

In the trace reconstruction problem a length-n string x yields a collection of noisy traces, where each is independently obtained from x by passing through a deletion channel, which deletes every symbol with some fixed probability. The main goal under this paradigm is to determine the required minimum number of i.i.d traces in order to reconstruct x with high probability. The focus of this work is to extend this problem to the model where each trace is a result of x passing through a deletion-insertion-substitution channel.