dehashSE: deanonymize (to some degree) a SummarizedExperiment post...

In trichelab/rehash: Lightly anonymize and deanonymize datasets for quasi-public distribution

Description

Verifiable de-anonymization requires metadata: a sample map, a feature map, a salting phrase, and knowledge of whether the salt phrase was applied to features, samples, both, or neither. If the metadata is left in the metadata slot of the dehydrated SummarizedExperiment, it will be used, else it must be supplied by the user. Optionally, the data itself may be verified.

Usage

dehashSE(x, meta = NULL, covs = NULL, check = FALSE)

Arguments

x	a RangedSummarizedExperiment to deanonymize, or a list*
meta	the meta-metadata required to reverse the hashes
covs	a data.frame of covariates (rows are samples, columns covs)
check	check assay row and column hashes? (FALSE; can be very slow)

Details

• A list result from rehash(x, ..., strip=TRUE) can be passed as x (the first argument to dehashSE) for testing purposes. In such cases, it will be disassembled into the necessary symbols to dehash the object. However, saving the list for distribution defeats the purpose of rehash()ing it, so don't do that in production. This hook exists ONLY for testing! If a list is detected as the argument, row and column hashes WILL be checked (regardless of the value of check in the function call), partly to discourage routine use of this functionality in production.

Value

   an object of the same class as x, but with cleartext dimnames

trichelab/rehash documentation built on Nov. 5, 2019, 10:58 a.m.