Metadata-conscious anonymous messaging
Authors: Giulia Fanti, Peter Kairouz, Sewoong Oh, Kannan Ramchandran, Pramod Viswanath
ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We further demonstrate empirically that adaptive diffusion hides the source effectively on real social networks. |
| Researcher Affiliation | Academia | University of Illinois at Urbana-Champaign, Champaign, IL 61801 University of California, Berkeley, CA 94701 |
| Pseudocode | Yes | Algorithm 1 Tree protocol (Fanti et al., 2015) and Algorithm 2 ML Source Estimator for Algorithm 1 |
| Open Source Code | No | The paper does not provide an explicit statement about releasing its source code or a link to a code repository for the methodology described. |
| Open Datasets | Yes | We evaluate diffusion with the first-spy estimator, and adaptive diffusion with a modification of the ML estimator in Proposition 4.1 that accounts for graph cycles. Figure 4 lists the probability of detection averaged over 200 trials, for p 0.15 (at its height, the Stasi employed 11 percent of the population as spies (Koehler, 1999)). Adaptive diffusion hides the source better than diffusion, and its probability of detection is close to the fundamental lower bound of p. This is likely because the mean node degree in the dataset is 25, so high-degree asymptotics are significant. While adaptive diffusion does not reach all nodes on a tree, cycles in the Facebook graph allow it to reach 81% of nodes within 20 timesteps. |
| Dataset Splits | No | The paper mentions simulating on a dataset but does not explicitly describe train/validation/test splits, percentages, or the methodology for partitioning data for these purposes. |
| Hardware Specification | No | The paper does not specify any hardware details (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not list specific version numbers for any software components, libraries, or solvers used in the experiments. |
| Experiment Setup | Yes | In the diffusion model used to generate Figure 1, each node propagates the message to each of its neighbors independently with probability q = 0.7 in each time step. We simulate adaptive and regular diffusion for q {0.1, 0.5}. |