Privacy Attacks on Schedule-Driven Data

Authors: Stephan A. Fahrenkrog-Petersen, Arik Senderovich, Alexandra Tichauer, Ali Kaan Tutak, J. Christopher Beck, Matthias Weidlich

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental An empirical evaluation for synthetic scheduling problems shows the effectiveness of informed privacy attacks and compares the results to theoretical bounds on uninformed attacks. Experiments with synthetic schedules indicate that informed privacy attacks indeed pose a threat and allow adversaries to make inference on private attributes of jobs. Figure 2: Upper and lower bounds for the expectation of total privacy loss (TPL) for an uninformed attack and a boxplot of TPL for an informed attack (when s is considered).
Researcher Affiliation Academia 1Humboldt-Universit at zu Berlin, Unter den Linden 6, 10117 Berlin, Germany 2 York University, 4700 Keele St, Toronto, ON M3J 1P3, Canada 3 University of Toronto, 5 King s College Rd, Toronto ON M5S 3G8, Canada
Pseudocode No The paper describes a dynamic programming procedure in the text for finding a single weight vector ('The problem of finding a single weight vector w1, . . . , wn can be solved via the following dynamic programming procedure.'), but it is not presented as a clearly labeled pseudocode block, figure, or algorithm.
Open Source Code Yes 1https://github.com/samadeusfp/aaai2023 schedule privacy
Open Datasets No The paper uses 'synthetically generated schedules' and does not mention or provide access to a public dataset. For example, 'An empirical evaluation for synthetic scheduling problems shows the effectiveness of informed privacy attacks... For each schedule, we solve the ISP by enumerating all possible solutions...'
Dataset Splits No The paper discusses 'synthetic scheduling problems' and '10,000 synthetically generated schedules' but does not specify any training, validation, or test splits. It directly computes TPL for these generated schedules.
Hardware Specification No The paper does not provide any specific hardware specifications used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python 3.8, PyTorch 1.9, etc.).
Experiment Setup No The paper describes the specific problem instances used in the motivating example (e.g., '3 patients having processing times p = (5, 3, 1), published permutation σ = (1, 2, 3), and weight domain of Xw = {1, . . . , 5}') and the general setup for generating synthetic schedules ('10,000 synthetically generated schedules'). However, it does not provide general experimental setup details such as hyperparameters, optimization settings, or other system-level configurations typically found in experimental setups for machine learning models.