Bio-inspired Information and Communication Technologies by Unknown

Author:Unknown
Language: eng
Format: epub
ISBN: 9783030242022
Publisher: Springer International Publishing

representing a specific machine learning algorithm, while has the inputs representing the medical data for patient . The set of known functions, are expected to predict a diagnosis or prognosis for x, for example the chances of patient x having Parkinson’s, cancer, heart attack, etc. The difference between this second scenario and the first outlined scenario is that in this scenario we assume that F is a known function (e.g., we know how to calculate Glasgow Coma Scale (GCS) using eye, verbal, and motor responses). Our goal is to apply F on both encrypted and unencrypted data and show that the output of a known function evaluation on both sets of data is the same. Outputs of these functions could be “health metrics”, “severity scores” and other clinical functions typically computed as a linear combination of privacy-protected factors, such as quantitative clinical or physiological patient data with a set of weights (coefficients). In practice, these functions are often designed to receive integer values as input variables and produce an integer as an output score. There are a large number of such studies conducted for diseases such as diabetes [33]. In many such modeling tasks, particularly when designing these models as commercial products, it is highly desirable to design, test and validate the functions privately. This scenario, in part, is simulated in this paper, where we apply known classification functions to encrypted and unencrypted data.

In very general terms, a potential exchange between the hospital and a center goes as follows. encrypts data, , of a particular patient x with E and sends the encrypted values to . applies the private function F to the encrypted data, thereby computing . The result of this computation, by the fully homomorphic property of E, is equal to . Next sends this result to , who decrypts the value and thus recovers . then sends this decrypted value back to . Based on the received evaluation of F, the final plaintext message, the hospital has a diagnosis or risk assessment for patient x. Thus, never learns the plaintext patient data x and never learns the function F, but does learn F(x).

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.