Differential privacy fails on fine-grained analysis like genome-wide association studies
Kurt Rohloff · Computing On Encrypted Data // Kurt Rohloff, Duality (Firstmark's Data Driven NYC) · Oct 22, 2019 · at 15:37
Dr. Kurt Rohloff, CTO and co-founder of Duality Technologies, explains the technical limitations of differential privacy during a fireside chat at Data Driven NYC.
0:00 / 0:07exact quote · 7.1s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“The challenge with that, it actually doesn't work very well if you need really fine-grained analysis as you would, for example genome-wide association studies.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Kurt Rohloff
Duality runs sub-second genome-wide association studies on encrypted NIH data
“We're actually at a point now where we're actually encrypting genomic data for NIH and running full-on genome-wide association studies on encrypted genomic data with, ah, sub-second response times.”
Homomorphic encryption is on the cusp of solving major privacy issues
“And the core technology that we have is something that's called homomorph encryption and it's been growing in, in many different spaces and a diff, couple different areas, but it's enabling and on the cusp of addressing all these very, kind of, serious issues …”
Lattice cryptography adoption is hindered by its unique compute model
“One of the reasons that so many, so few people have been able to actually use this technology so far is that it's a very, very different compute model.”
Duality Technologies is fundamentally a data science company, not just security
“Though ostensibly duality technologies is really about security and cryptography, we're actually very much heart and soul at, a data science company, and how we can enable analytics on data that is protected and sensitive.”