Communications of the ACM
Secure Multiparty Computation
Credit: Weibel Christophe / Shutterstock
Distributed computing considers the scenario where a number of distinct, yet connected, computing devices (or parties) wish to carry out a joint computation of some function. For example, these devices may be servers that hold a distributed database system, and the function to be computed may be a database update of some kind. The aim of secure multiparty computation is to enable parties to carry out such distributed computing tasks in a secure manner. Whereas distributed computing often deals with questions of computing under the threat of machine crashes and other inadvertent faults, secure multiparty computation is concerned with the possibility of deliberately malicious behavior by some adversarial entity (these have also been considered in the distributed literature where they are called Byzantine faults). That is, it is assumed that a protocol execution may come under "attack" by an external entity, or even by a subset of the participating parties. The aim of this attack may be to learn private information or cause the result of the computation to be incorrect. Thus, two important requirements on any secure computation protocols are privacy and correctness. The privacy requirement states that nothing should be learned beyond what is absolutely necessary; more exactly, parties should learn their output and nothing else. The correctness requirement states that each party should receive its correct output. Therefore, the adversary must not be able to cause the result of the computation to deviate from the function that the parties had set out to compute.
Key Insights
Secure multiparty computation can be used to solve a wide variety of problems, enabling the utilisation of data without compromising privacy. Consider, for example, the problem of comparing a person's DNA against a database of cancer patients' DNA, with the goal of finding if the person is in a high risk group for a certain type of cancer. Such a task clearly has important health and societal benefits. However, DNA information is highly sensitive, and should not be revealed to private organizations. This dilemma can be solved by running a secure multiparty computation that reveals only the category of cancer that the person's DNA is close to (or none). In this example, the privacy requirement ensures that only the category of cancer is revealed, and nothing else about anyone's DNA (neither the DNA of the person being compared nor the DNA of the patients in the database). Furthermore, the correctness requirement guarantees that a malicious party cannot change the result (for example, make the person think that they are at risk of a type of cancer, and therefore need screening).
No entries found