Protecting privacy in genomic databases

Argh. The amount of headache we must go through to lock down this information is ridiculous compared to the risk. While this might be an issue for Bill Clinton or Kanye West, the vast majority of people won't be terribly adversely affected by breaches of their medical records.

A much better solution than encrypting every byte of data and vainly trying to secure the impossible (to say nothing of purposely corrupting the data) is just step up enforcement: punish people who leak, sell, and buy private health information more assiduously. Data breaches will happen; the social contract is our best defense against them.

I've worked with Biobankers/Bioinformaticians on this type of problem before and none of them wanted us to got the Differential Privacy route. Why? The vast majority of genomic diseases are rare diseases (with 10s-1000s of carriers). Destroying information is not an option with small sample sizes. For huge GWAS studies, it could be an option. But in the general case, security is preferable over privacy when a choice needs to be made.

No scientist will ever use a database like that and no government agency will fund it. Data provenance is absolutely vital for getting your work published and good data accessible is infinitely more valuable than garbage accessible to everyone. I say this as someone who is dedicated to open data and open science. Privacy is something that science has to deal with. There may be a set of questions that can still be answered with degraded data, but I bet that if such a system were built scientists would still try to used the degraded data to answer questions which it could not answer, wasting their time and federal money.

So they want to improve privacy by corrupting data...seems like a bad solution.