Securing the Cloud

The potential benefits of cloud computing are amazing. Imagine for example your medical records being in the cloud, accessible immediately from anywhere in an emergency...

But wait a minute. What if the cloud is not secure? Already there have been a couple of instances where a cloud was hacked, or where there was a sustained loss of service. The cloud represents a single point of failure, and thus must not be allowed to fail. And what about privacy - you don't want anyone and everyone poking around in your medical records. The solution here is encryption, but a particularly sophisticated type of encryption. For example you might want your medical records to be completely available to you and your surgeon, and partially available to your GP and nurse, and somewhat available to your medical insurer. What's wanted in fact is known as "Attribute-Based Cryptography", where your access to the data depends on who you are and what attributes you have.

Well cometh the hour cometh the solution. As often happens in research, in a completely seperate development from cloud computing, mathematicians have recently come up with the perfect solution to the problem of attribute-based cryptography, based on the mathematics of "pairing-based cryptography". Previous efforts floundered on the problem of "collusion attacks". Say one needs attribute A and attribute B to access certain data. I have attribute A only, you have attribute B only, each of us individually does not have sufficient attributes to access the data, but if the two of us get together... The new methods prevents this. This is a very hot research area.

If you want to read more, see

http://eprint.iacr.org/2010/565

and

http://ece.wpi.edu/~mingli/papers/Li_SecureComm10.pdf

But academics having solved a problem do not linger long on it. There are other problems of interest. Unfortunately commercial exploitation of such ideas is often delayed, as knowledge of these new techniques is not widespread.
Here is another problem of particular interest to an Irish audience - Electronic Voting. We in Ireland dabbled with it before but the proposed method then was totally insecure. Ideally we would like to encrypt each vote and yet still be able to count it. However normally you can't do maths (like counting) on encrypted data. The solution is "homomorphic cryptography". Google for it. We don't quite have a perfect solution for it yet, but we are getting there... Another fascinating and hyperactive research area.