Insider Attack Game

Aggarwal, P., Cranford, E.A., Tambe, M., Lebiere, C., Gonzalez, C. (2023). Deceptive Signaling: Understanding Human Behavior Against Signaling Algorithms. In: Bao, T., Tambe, M., Wang, C. (eds) Cyber Deception. Advances in Information Security, vol 89. Springer, Cham. https://doi-org.cmu.idm.oclc.org/10.1007/978-3-031-16613-6_5

The Insider Attack Game is a simulated real-world scenario of a network composed of six nodes (computers) with resource constranined alocation where a defender can only protect two nodes. The game evolves over four phases, each with 25 decisions. Each decision comprises two stages in which participants play the role of attackers. In the first stage of the decision, participants must choose a node to attack while defenders protect nodes (computers) by randomly allocating limited resources. In the second stage of the decision, the defender decides probabilistically to either warn the attacker that the node is defended (which could be a lie) or not provide such a warning, and then the attacker (participant) chooses to continue the attack or not.

Data files

The human experiment dataset is available on OSF: https://osf.io/c7ntu/?view_only=0e6261b5e818440495d9917044611758

For this particular implementation, we are using:

- 2022-MURIBookChapter-FullData-IAG.csv
- AllConditions_humanData_dictionary.txt (dictionary with variables information)

max_decays_IAG.csv was generated after running Model-Fitting for each model containing the best decay for each participant.

NOTE: For convenience, we have added the CSV files to the Data folder.

A tour of the code

• insiderAttackGame.py contains the implementation of the IBL model for Model-Tracing.

  • To run this code, you will need the two files located inside the Data folder at the project's root. One of the files can be found in the OSF link, and the other is automatically generated after running the IAG.Rmd script.
  • This script will generate automatically the results of the IBL models in a file named IAG_Tracing_Data_Personalized.csv recorded in the folders Generated_Models_Data\Tracing_Results on the root of the project.

• runFitting.py has the development of the Model-Fitting implementation according to the IBL model in the insiderAttackGame.py.

  • In this script, each participant's data will be run by 291 IBL models, and each model will have a decay value in the range of [0.1 - 3] with increments of 0.01.
  • To run this code, you will need the file 2022-MURIBookChapter-FullData-IAG.csv located in the Data folder.
  • This script will generate automatically the results of all the models per participant and record it as Fitting_Tracing_Data.csv in the folders Generated_Models_Data\Fitting_Tracing_Results on the root of the project.

• confusionMatrix.py contains functions to generate confusion matrixes and learning curves in different settings.

  • To run this code, you will need the file IAG_Tracing_Data_Personalized.csv generated by the insiderAttackGame.py script available in the folder Generated_Models_Data\Tracing_Results.
  • This script will automatically generate images in the PNG format with confusion matrixes and learning curves of the Model Predictions and/or the Human Choices in different settings. These images will be automatically stored in a folder named Plots_{path}\Individual at the project's root. {path} is replaced by either ConfusionMatrixes or LearningCurves depending on the settings analyzed.

• IAG.Rmd is an R code script used to assess the best fitting decay for each participant based on the 291 Models generated.

  • The best fitting is assessed based on the highest Synchronization Rates (SyncRate) of each model for each participant. If multiple models have the same SyncRate, the model with the highest decay is selected for that participant.
    • SyncRate examines the synchronization between the model prediction and each human choice. We determined whether the model prediction was the same as the actual human action for each decision. If it was the same, the synchronization value for that decision was 1; otherwise, it was 0.
    • We calculated the SyncRate average per participant for each IBL model (with a different decay value) for the last 50 decisions (the model only predicts the last 50 decisions).
  • To run this script, you must first run the runFitting.py script and use the generated Fitting_Tracing_Data.csv file located in Generated_Models_Data\Fitting_Tracing_Results.
  • This script will automatically generate the max_decays_IAG.csv in the same folder where the script is located (same file as the one in the Data folder).

  NOTE: The R script was run in Rstudio 2024.04.2+764 and needs an independent project. A copy of the file Fitting_Tracing_Data.csv needs to be added inside a folder named Data\IAG_Game in the root of the R project. After running this script, the generated file needs to be copied into the Python project in the Data folder.

NOTE: The scripts also automatically create all folders where generated files are recorded; manually adding these folders to the project is unnecessary.

Installation

Ensure Python version 3.8 or later is installed.

Run python3 -m venv venv or python -m venv venv to create a virtual environment.

Run source venv/bin/activate or .\venv\Scripts\activate to activate the virtual environment.

Run pip install -r requirements.txt to install the required Python packages into the virtual environment.

NOTE: If you do not want to create a virtual environment, just run pip install -r requirements.txt.

More information

For more information or details, please contact [email protected].