Manifesto for secure data sharing

Last year in a project meeting in Dagstuhl, Germany I explained complex security scenarios for remote sharing of databases among multiple teams. My peer Breanndan from the University of Amsterdam commented: 'Why did you make it so complex, when we need it simple? Look: I have a database, my colleague has even more data, and you have the data mining software. We want to run your software over our data, that’s it! Optimally, we need to form a dynamic secure data sharing group on-the-fly. We give each other the access rights, we run the job, and then we immediately disband the group. '

'Yes - I replied – this is precisely what my Virtual Organizations are for. And by the way, this isn’t a new idea.' But his point was well taken and it got me thinking. This "simple" scenario is in fact complex to implement. But really, why did we make it so complex for the end user?

To step back a bit, over a decade ago people understood the need to share large data, applications, and even machines. Much before the pervasive Web 2.0 and P2P file sharing craze, this need has been acknowledged by the science community. There, the idea of the Virtual Organizations (VO) was born: dynamic groups of mutual trust, that people (or processes) could set up on the fly. Entities in a VO would share a digital identity scheme, to securely identify and authorize each other, share whatever, and then – disband the VO as soon as the task was complete. Soon, the first Virtual Organizations were implemented… or were they really?

Nope.  Bureaucracy won. It is striking how unfriendly to the user the administrative policies could become. In corporations and universities alike, there is one security bottleneck: the human administrator with his procedures. On the pretext of technology requirements, in the fear of responsibility, we have killed productivity and forgotten the premise of on-the-fly, secure yet easy sharing.

I have not seen many production virtual organization in action. Instead, I saw a medical project where thousands of anonymized patient records, in the absence of effective sharing mechanism, were shipped to another institution on a hard drive in a parcel. I have also been to a research center where a crafty project team has quietly set up their own wifi router with a direct Internet cable, to bypass the absurd security imposed by their institution.

Is this rational? In the case of mission-critical data, I agree. But we’ve grown mature enough to understand that there are various degrees of confidentiality in various data. In many cases imposing equally strict security regime on all data is ridiculous. Hence my…

Manifesto for secure data sharing

1. Free the ordinary users! Let them decide to share.

Today, institutions leave this decision to an expert or an administrator. But that’s not practical. Everyone owns data these days. Sharing data is not an IT concept any more, it is our daily bread. Web 2.0 and ubiquitous P2P file shares, whether we want them or not, demonstrate it best. So: administrators should guard the data critical for the enterprise. Temporary data of a project should be owned and managed directly by the team.

2. User-friendly data sharing interface should not require IT skills.

Because sharing is not an expert action, it should not need an expert’s involvement. Optimally, it should be a drag-and-drop GUI.

3. Data sharing must be easy, efficient and take seconds.

I know cases when the procedure to add members to a group can take hours or even days, mainly due to the involvement of third parties. I think multiple levels of human acceptance are okay, if that really is the policy of an institution in case of critical data. Technology should enable, but not enforce such model.

Breanndan’s remark last year was indeed very stimulating. We got back to the drawing board. And we came up with the entirely new way of thinking of the data sharing security, which was basically what the users wanted. This was baptized AdHoc, as it empowers ordinary users ad-hoc collaborations in three seconds; just like the scenario described at the beginning of this article. AdHoc 1.1.0 has just been announced a few days ago. My skilled team crafted this hilarious movie  which describes AdHoc with tongue in cheek.

Big kudos to the poor virolabers including myself who became overnight movie stars at one project retreat. Disclaimer: any resemblance of the BigDisaster story to the existing numerous, successful collaborative data sharing projects is purely coincidental.

While AdHoc is now available for users, I’m also interested in further discussion on practical requirements for secure yet efficient and easy data sharing.  Feel free to comment.

Comments

[a.k.]

AdHoc, great idea.

[Chris Wilk]

Though it is related to other technologies (clouds) the newly established Google's effort "The Data Liberation Front" http://www.dataliberation.org/ has quite a similar goal.

[Miha Ahronovitz]

I am watching in wonder the transformation of Gridwisetech into a database expertise center from a grid expertise center. Not only database, but an Oracle Database. The two kinds of expertise are rare to find into a single individual.

I am dreaming or not...

Miha

[Pawel Plaszczak]

Thanks for the kind comment Miha. In fact, this transformation came naturally. We deployed grids in corporations to make their IT scalable, on-demand and agile. But in many cases scaling out the processing was of no use, as long as the monolithic data i/o was on the way.

Now that we have built expertise in scaling out both the database and the processing, it is easier to tackle scalability problems of an enterprise holistically.