Injecting Undetectable Backdoors in Obfuscated Neural Networks and Language Models

Authors: Alkis Kalavasis, Amin Karbasi, Argyris Oikonomou, Katerina Sotiraki, Grigoris Velegkas, Manolis Zampetakis

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Our work is theoretical.
Researcher Affiliation Academia Alkis Kalavasis Yale University alkis.kalavasis@yale.edu Amin Karbasi Yale University amin.karbasi@yale.edu Argyris Oikonomou Yale University argyris.oikonomou@yale.edu Katerina Sotiraki Yale University katerina.sotiraki@yale.edu Grigoris Velegkas Yale University grigoris.velegkas@yale.edu Manolis Zampetakis Yale University manolis.zampetakis@yale.edu
Pseudocode No The paper describes procedures in descriptive text and numbered steps (e.g., Section 5), but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code No The paper is theoretical and does not mention releasing any source code. The NeurIPS Paper Checklist explicitly marks 'NA' for questions related to code availability.
Open Datasets No The paper is theoretical and does not conduct experiments with specific datasets. It provides a generic definition of a dataset 'S = {(xi, yi)}m i=1' for its theoretical framework, but no concrete dataset is specified or made available.
Dataset Splits No The paper is theoretical and does not describe experimental validation or specific dataset splits (training, validation, testing) for empirical results.
Hardware Specification No The paper is theoretical and does not mention any specific hardware used for experiments.
Software Dependencies No The paper is theoretical and does not list any specific software dependencies with version numbers.
Experiment Setup No The paper is theoretical and does not describe an experimental setup with specific hyperparameters or system-level training settings.