Turvalise ühisarvutuse protokollid kõrgtaseme programmeerimiskeelest
Date
2014
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Turvalise ühisarvutuse abil on võimalik sooritada privaatsust säilitavaid arvutusi mitmelt osapoolelt kogutud andmetega.
Tänapäeva digitaalses maailmas on andmete konfidentsiaalsuse tagamine üha raskemini teostatav.
Turvalise ühisarvutuse meetodid nagu ühissalastus ja Yao sogastatud loogikaskeemid võimaldavad teostada privaatsust säilitavaid arvutusprotokolle,
mis ei lekita konfidentsiaalseid sisendandmeid. Aditiivne ühissalastuse skeem on väga efektiivne algebraliste ringide tehete sooritamiseks
fikseeritud bitilaiusega andmetüüpide peal. Samas on seda kasutades raske ehitada protokolle, mis nõuavad paindlikumaid bititaseme operatsioone.
Yao sogastatud loogikaskeemide meetod töötab aga igasuguse bitilaiusega andmete peal ja võimaldab väärtustada mistahes Boole'i funktsioone.
Neid kahte meetodit koos kasutades ehitame turvalise hübriidprotokolli, mis kujutab endast üldist meetodit privaatsust säilitavate arvutuste teostamiseks
bitikaupa ühissalastatud andmete peal. Loogikaskeeme vajalikeks arvutusteks on lihtne saada kahe kaasaegse turvalise ühisarvutuse jaoks mõeldud
kompilaatori abil, mis muundavad C programmi loogikaskeemiks --- PCF ja CBMC-GC. Meie hübriidprotokolli prototüüp privaatsust säilitaval arvutusplatvormil Sharemind
saavutab praktilisi jõudlustulemusi, mis on võrreldavad teiste kaasaegsete lahendustega. Lisaks kahe osapoolega arvutustele pakub meie prototüüp võimekust teostada mitmekesiseid arvutusi
üldises turvalise ühisarvutuse arvutusmudelis.
Hübriidprotokoll ja loogikaskeemide kompilaatorid võimaldavad koos kasutades lihtsalt ja efektiivselt luua üldkasutatavaid turvalise ühisarvutuse protokolle
mistahes Boole'i funktsioonide väärtustamiseks.
Secure multi-party computation (SMC) enables privacy-preserving computations on data originating from a number of parties. In today's digital world, data privacy is increasingly more difficult to provide. With SMC methods like secret sharing and Yao's garbled circuits, it is possible to build privacy-preserving computational protocols that do not leak confidential inputs to other parties. The additive secret sharing scheme is very efficient for algebraic ring operations on fixed bit-length data types. However, it is difficult to build protocols that require robust bit-level manipulation. Yao's garbled circuits approach, in contrast, works on arbitrary bit-length data and allows the evaluation of any Boolean function. Combining the two methods, we build a secure hybrid protocol, which provides a general method for building arbitrary secure computations on bitwise secret-shared data. We are able to generate circuits for the protocol easily by using two state-of-the-art C to circuit compilers designed for SMC applications --- PCF and CBMC-GC. Our hybrid protocol prototype on the Sharemind privacy-preserving computational platform achieves practical performance comparable to other recent work. In addition to two-party computations, our prototype provides the ability to perform a set of diverse computations in a generic SMC computational model. The hybrid protocol together with the circuit compilers provides a simple and efficient toolchain to build general-purpose SMC protocols for evaluating any Boolean function.
Secure multi-party computation (SMC) enables privacy-preserving computations on data originating from a number of parties. In today's digital world, data privacy is increasingly more difficult to provide. With SMC methods like secret sharing and Yao's garbled circuits, it is possible to build privacy-preserving computational protocols that do not leak confidential inputs to other parties. The additive secret sharing scheme is very efficient for algebraic ring operations on fixed bit-length data types. However, it is difficult to build protocols that require robust bit-level manipulation. Yao's garbled circuits approach, in contrast, works on arbitrary bit-length data and allows the evaluation of any Boolean function. Combining the two methods, we build a secure hybrid protocol, which provides a general method for building arbitrary secure computations on bitwise secret-shared data. We are able to generate circuits for the protocol easily by using two state-of-the-art C to circuit compilers designed for SMC applications --- PCF and CBMC-GC. Our hybrid protocol prototype on the Sharemind privacy-preserving computational platform achieves practical performance comparable to other recent work. In addition to two-party computations, our prototype provides the ability to perform a set of diverse computations in a generic SMC computational model. The hybrid protocol together with the circuit compilers provides a simple and efficient toolchain to build general-purpose SMC protocols for evaluating any Boolean function.