Private Computing with Garbled Circuits [Applications Corner]
Riccardo Lazzeretti, Mauro Barni · IEEE Signal Processing Magazine · 2013
Rapid technological advances in multimedia processing, healthcare, and biometrics, among others, have given rise to a variety of new signal processing applications for which security aspects can no longer be dealt with by classical cryptographic methods. The classical security model is targeted toward protecting the communication of two trusted parties against a potentially malevolent third party. In such cases, it is sufficient that secure cryptographic primitives are applied on top of transmission and processing modules. In an increasing number of applications, however, the classical security model is no longer adequate since one or multiple of the parties involved in the communication, distribution, or processing of the data may not be trusted. For example, a database server may not be trusted, creating the need for methods that hide the content of query to the database server, while still allowing the query to be resolved. As another example, we may consider a remote diagnosis service, where a nontrusted party is tasked with processing sensitive medical data. In such a scenario, we need methods and technologies to carry out the task (medical diagnosis) without leaking the private data of the patient (including the diagnosis results).