A Privacy Preserving Collusion Secure DCOP Algorithm
Authors: Tamir Tassa, Tal Grinshpoun, Avishai Yanay
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate its performance on different benchmarks, problem sizes, and constraint densities. We show that achieving security against coalitions is feasible. ... We begin by evaluating the runtime of the compare CPA cost to upper bound sub-protocol, which is a central and computationally expensive part of PC-Sync BB. ... Now we turn to the runtime performance evaluation of the full PC-Sync BB algorithm. ... The algorithms were implemented and executed in the Agent Zero simulator [Lutati et al., 2014], running on a hardware comprised of an Intel i7-6820HQ processor and 32GB memory... |
| Researcher Affiliation | Academia | Tamir Tassa1 , Tal Grinshpoun2 and Avishay Yanai3 1The Open University, Ra anana, Israel 2Ariel University, Ariel, Israel 3Bar-Ilan University, Ramat Gan, Israel tamirta@openu.ac.il, talgr@ariel.ac.il, ay.yanay@gmail.com |
| Pseudocode | Yes | Algorithm 1 PC-Sync BB (executed by agent Ak) |
| Open Source Code | Yes | For efficiency and reproducibility we used the original implementation of Ben-Efraim and Omri [2017].3 The executions were over LAN with EC2 machines of type c5.large in Amazon s North Virginia data center with every agent running on a separate machine. ... 3https://github.com/cryptobiu/Protocols/tree/master/Concrete Efficiency Improvements to Multiparty Garbling with an Honest Majority |
| Open Datasets | No | The paper mentions generating DCOPs (randomly generated DCOPs, distributed 3-color graph coloring problems, scale-free networks) but does not provide access information or formal citations for public datasets that were used. The problem instances are generated based on parameters, but no existing public dataset access is given. For example, "The first benchmark consists of unstructured randomly generated DCOPs..." |
| Dataset Splits | No | The paper does not explicitly provide details about training/validation/test dataset splits. It mentions using "50 problem instances (for each setting/benchmark)" but not how these instances were split for training, validation, or testing. |
| Hardware Specification | Yes | The executions were over LAN with EC2 machines of type c5.large in Amazon s North Virginia data center with every agent running on a separate machine. ... The algorithms were implemented and executed in the Agent Zero simulator [Lutati et al., 2014], running on a hardware comprised of an Intel i7-6820HQ processor and 32GB memory... |
| Software Dependencies | Yes | The algorithms were implemented and executed in the Agent Zero simulator [Lutati et al., 2014]... For efficiency and reproducibility we used the original implementation of Ben-Efraim and Omri [2017]. |
| Experiment Setup | Yes | We measured performance for various values of n, where q (the maximum value of a single binary constraint) is set to 100. Hence, as the maximum cost of any solution is q := n 2 q + 1 and S is set to be the smallest power of 2 greater than 2q , then n fully determines ℓ= log S and, consequently, also the size of the circuit C that the protocol uses. ... In the first experiment, presented in Figure 1, we fix the number of agents to n = 7 and the domain sizes to 6, and vary the constraint density 0.3 p1 0.9. ... In the second experiment, shown in Figure 2, we fix the constraint density to p1 = 0.3 and the domain sizes to 6, and vary the number of agents 5 n 9. Here and in the following scalability experiments we use a cutoff time of 30 minutes for online PC-Sync BB. |