Configuration

Settings

• image : The docker image to use for bunnyfinder service.
• dbconnect: The database connection string, current only support mysql database, e.g. root:password@tcp(ip:port)/dbname
• min_malicious_idx: The min index that actor as a malicious validator.
• max_malicious_idx: The max index that actor as a malicious validator.
• duration_per_strategy: The execute duration for each strategy in minutes, this is used when strategy is set to all or multiple strategies.
• strategy: The strategy to use, all supported strategy is listed in strategy description.
• openai_key: The openai api key, this is used when strategy is set to ai, also supported DeepSeek api.
• openai_base_url: The openai base url, this is used when strategy is set to ai, also supported DeepSeek api.
• llm_model: The large language model to use, this is used when strategy is set to ai, We recommend using deepseek-r1, deepseek-v3, gpt-4o.
• replay_project: The project uuid to replay the attack strategies from the database, this is used when strategy is set to replay.

Strategy Description

• simple: Execute a simple attack strategy, which is a random attack strategy without any profile.
• random: Randomly generate an attack strategy with a random profile.
• exante: Execute ex-ante attack.
• ext_exante: Execute ex-ante attack with some additional random attack strategies.
• sandwich: Execute sandwich attack.
• ext_sandwich: Execute sandwich attack with some additional random attack strategies.
• staircase: Execute staircase attack.
• ext_staircase: Execute staircase attack with some additional random attack strategies.
• unrealized: Execute unrealized attack.
• ext_unrealized: Execute unrealized attack with some additional random attack strategies.
• withholding: Execute withholding attack.
• ext_withholding: Execute withholding attack with some additional random attack strategies.
• ai: Use AI to generate attack strategies.
• replay: Replay the attack strategies from the database, this is used for testing and debugging.
• syncwrong: Execute a sync attack.
• all: Execute all existing strategies, switch to a different existing strategy randomly every duration_per_strategy minutes.

Example

Bellow is an example of the configuration file, it use the tscel/bunnyfinder:stable docker image, connect to a local mysql database, and execute the random and exante strategies with a duration of 30 minutes for each strategy.

  image: 'tscel/bunnyfinder:stable'
  strategy: 'random,exante'
  dbconnect: 'eth:12345678@tcp(127.0.0.1:3306)/eth'
  max_malicious_idx: "85"
  duration_per_strategy: "30"

Below is an example of the configuration file, using the deepseek-r1 model provided by aliyun, the tscel/bunnyfinder:stable docker image, connecting to a local MySQL database, and executing the ai strategy.

  image: 'tscel/bunnyfinder:stable'
  strategy: 'ai'
  dbconnect: 'eth:12345678@tcp(127.0.0.1:3306)/eth'
  max_malicious_idx: "85"
  openai_key: "sk-xxxxxxxxx"
  openai_base_url: "https://dashscope.aliyuncs.com/compatible-mode/v1"
  llm_model: "deepseek-r1"

Below is an example of the configuration file, connect to a local MySQL database, and execute the replay strategy to replay the attack strategies from the database.

  image: 'tscel/bunnyfinder:stable'
  strategy: 'replay'
  dbconnect: 'eth:12345678@tcp(127.0.0.1:3306)/eth'
  max_malicious_idx: "85"
  replay_project: "9c344c56-573b-459e-b0a8-952301c9e294"