Implementing Privacy Preserving Auction Protocols (Master Thesis Final Talk)
Total Page:16
File Type:pdf, Size:1020Kb
Chair of Network Architectures and Services Department of Informatics Technical University of Munich Implementing Privacy Preserving Auction Protocols (Master Thesis Final Talk) Markus Teich, B. Sc. [email protected] March 23, 2017 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Chair of Network Architectures and Services Department of Informatics Technical University of Munich Motivation In many existing auction systems sellers can: • Abuse loosing bid information • Collude with bidder(s) and change the outcome Our goals are: • Reduce requirement of trust in auction platform operators • Create usably fast system M. Teich — Auctions 2 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Related Work Cryptographic auction protocols provide: • Correctness of the outcome • Non-Repudiation • Various degrees of privacy We choose the work of Felix Brandt because: • No trusted third party required • Beneficial privacy properties • Reviewed M. Teich — Auctions 3 first price first price private outcome public outcome M + 1st price M + 1st price private outcome public outcome Chair of Network Architectures and Services Department of Informatics Technical University of Munich Auction Formats Sealed bid vs. Incrementing M. Teich — Auctions 4 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Auction Formats Sealed bid vs. Incrementing first price first price private outcome public outcome M + 1st price M + 1st price private outcome public outcome M. Teich — Auctions 4 Chair of Network Architectures and Services Department of Informatics Technical University of Munich libbrandt Overview • git://gnunet.org/libbrandt • Auction Outcome determination (all four formats) • Algorithms by Felix Brandt translated to Ed25519 • Around 4000 lines of C • Depends on libgcrypt v1.7+ and libgnunetutil v13.0+ • GPLv3+ M. Teich — Auctions 5 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Research Questions • How does the architecture look like? • How much do we improve performance by using Ed25519? • How scalable is the system? M. Teich — Auctions 6 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Architecture Figure 1: Simplified System Architecture M. Teich — Auctions 7 Chair of Network Architectures and Services Department of Informatics Technical University of Munich GNUnet Auction ADF := auction description file • create: Initiate seller, output: ADF • join: Join auction as bidder, input: ADF • info: Retrieve status of running auctions, extract info from ADF M. Teich — Auctions 8 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Research Questions X How does the architecture look like? • How much do we improve performance by using Ed25519? • How scalable is the system? M. Teich — Auctions 9 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Measurement Setup • Lenovo X240 notebook computer • Core i7-4600U CPU (2.1GHz) • Single CPU core used • Generated 3072 bit RSA key with ssh-keygen • Bids chosen uniformly at random • Median of ∼ 10 iterations M. Teich — Auctions 10 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Measurement Results 1 Figure 2: First Price Private Outcome Auction with five Prices. M. Teich — Auctions 11 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Measurement Results 2 Figure 3: First Price Private Outcome Auction with five Bidders. M. Teich — Auctions 12 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Wassenberg Heuristic Figure 4: Possible (t, u) Pairs for M = 3, n = 7 and any k ≥ n. M. Teich — Auctions 13 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Measurement Results 3 Figure 5: M + 1st Price Private Outcome Auction with n = k = M + 2. M. Teich — Auctions 14 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Measurement Results 4 Figure 6: All libbrandt Algorithms with 512 Prices. M. Teich — Auctions 15 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Performance Conclusion Compared to RSA: • Computation cost reduced to 7% • Bandwidth cost reduced to 8% • Wassenberg Heuristic worth looking into M. Teich — Auctions 16 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Research Questions X How does the architecture look like? X How much do we improve performance by using Ed25519? • How scalable is the system? M. Teich — Auctions 17 Given a good multicast-implementation the systems scalability only de- pends on the network speed and computing power of the seller and the bidders. Chair of Network Architectures and Services Department of Informatics Technical University of Munich Scalability • Computations in O(n3) • Participating in several auctions on the same host is problematic. • Only Sellers and bidders involved → No need to scale a platform or auctioneer M. Teich — Auctions 18 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Scalability • Computations in O(n3) • Participating in several auctions on the same host is problematic. • Only Sellers and bidders involved → No need to scale a platform or auctioneer Given a good multicast-implementation the systems scalability only de- pends on the network speed and computing power of the seller and the bidders. M. Teich — Auctions 18 Depending on the auction format we need between 4kiB/s and 20kiB/s. → computation cost > bandwidth cost. Moore’s Law is faster than Nielsen’s Law. Chair of Network Architectures and Services Department of Informatics Technical University of Munich Computation Cost vs. Bandwidth Cost Which network speed leads to equal time consumption? M. Teich — Auctions 19 → computation cost > bandwidth cost. Moore’s Law is faster than Nielsen’s Law. Chair of Network Architectures and Services Department of Informatics Technical University of Munich Computation Cost vs. Bandwidth Cost Which network speed leads to equal time consumption? Depending on the auction format we need between 4kiB/s and 20kiB/s. M. Teich — Auctions 19 Moore’s Law is faster than Nielsen’s Law. Chair of Network Architectures and Services Department of Informatics Technical University of Munich Computation Cost vs. Bandwidth Cost Which network speed leads to equal time consumption? Depending on the auction format we need between 4kiB/s and 20kiB/s. → computation cost > bandwidth cost. M. Teich — Auctions 19 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Computation Cost vs. Bandwidth Cost Which network speed leads to equal time consumption? Depending on the auction format we need between 4kiB/s and 20kiB/s. → computation cost > bandwidth cost. Moore’s Law is faster than Nielsen’s Law. M. Teich — Auctions 19 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Research Questions X How does the architecture look like? X How much do we improve performance by using Ed25519? X How scalable is the system? M. Teich — Auctions 20 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Final Status • https://gnunet.org/libbrand-thesis • More measurement data • Detailed algorithm and protocol description • Speed suffices for real world usage • Architecture design M. Teich — Auctions 21 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Future Work • GNUnet Auction system (GSoC 2017) • Escrowed payment service using GNU Taler • Auction distribution platform • Price pool distribution • Reputation system for sellers M. Teich — Auctions 22 Chair of Network Architectures and Services Department of Informatics Technical University of Munich Closing Q&A M. Teich — Auctions 23.