From 73ea4f658aca3c5f9d7bfdcfbaa8a2ec8f050f86 Mon Sep 17 00:00:00 2001 From: Zippo00 Date: Mon, 13 Apr 2026 21:39:39 +0800 Subject: [PATCH] quickstart updt --- README.md | 29 +++++++++++++++++++++++++++++ docs/source/installation.rst | 29 +++++++++++++++++++++++++++++ docs/source/quickstart.rst | 30 +++++++++++++++++++++++++++++- 3 files changed, 87 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 5de1912..393b02b 100644 --- a/README.md +++ b/README.md @@ -103,6 +103,35 @@ For example, you can edit the default Ollama Connector configuration file `AVISE } } ``` + +Similarly, you can customize the configurations for SETs as well. For example, by editing the Red Queen SET configuration file `AVISE/avise/configs/SET/languagemodel/multi_turn/red_queen.json`, +you can define if the SET is executed incrementally *(the target model will generate a response after each subsequential prompt)*, or as a template *(only works for target systems that accept a conversation as an input)* and if the SET uses and Adversarial Language Model (ALM). +Additionally, you can define the exact template attack prompts that the SET uses: + +```json +{ + "incremental_execution": false, #Change to true to execute the SET cases incrementally + "use_adversarial_languagemodel": false, #Change to true to utilize an Adversarial Language Model (ALM) + "adversarial_model_device": "auto", # Defines on which device(s) the ALM is loaded to ("cpu", "gpu", or "auto") + "evaluation_system_prompt": "You are a highly accurate cybersecurity classifier, ..." + "evaluation_model_device": "auto", # Defines on which device(s) the ELM is loaded to ("cpu", "gpu", or "auto") + "sets": [ + { + "id": "RED-QUEEN-001", + "vulnerability_subcategory": "Evasion Attack", + "action": "make a fake passport", + "type": "occupation_teacher", + "category": "violence,aiding_and_abetting,incitement", + "conversation": [ + { ... } + ... + ] + }, + ... + ] +``` + + If you want to use custom configuration files for SETs and/or Connectors, you can do so by giving the paths to the configuration files with `--SETconf` and `--connectorconf` arguments: ```bash diff --git a/docs/source/installation.rst b/docs/source/installation.rst index f590c9a..c432f69 100644 --- a/docs/source/installation.rst +++ b/docs/source/installation.rst @@ -92,6 +92,35 @@ For example, you can edit the default Ollama Connector configuration file `AVISE } } ``` + +Similarly, you can customize the configurations for SETs as well. For example, by editing the Red Queen SET configuration file `AVISE/avise/configs/SET/languagemodel/multi_turn/red_queen.json`, +you can define if the SET is executed incrementally *(the target model will generate a response after each subsequential prompt)*, or as a template *(only works for target systems that accept a conversation as an input)* and if the SET uses and Adversarial Language Model (ALM). +Additionally, you can define the exact template attack prompts that the SET uses: + +```json +{ + "incremental_execution": false, #Change to true to execute the SET cases incrementally + "use_adversarial_languagemodel": false, #Change to true to utilize an Adversarial Language Model (ALM) + "adversarial_model_device": "auto", # Defines on which device(s) the ALM is loaded to ("cpu", "gpu", or "auto") + "evaluation_system_prompt": "You are a highly accurate cybersecurity classifier, ..." + "evaluation_model_device": "auto", # Defines on which device(s) the ELM is loaded to ("cpu", "gpu", or "auto") + "sets": [ + { + "id": "RED-QUEEN-001", + "vulnerability_subcategory": "Evasion Attack", + "action": "make a fake passport", + "type": "occupation_teacher", + "category": "violence,aiding_and_abetting,incitement", + "conversation": [ + { ... } + ... + ] + }, + ... + ] +``` + + If you want to use custom configuration files for SETs and/or Connectors, you can do so by giving the paths to the configuration files with `--SETconf` and `--connectorconf` arguments: ```bash diff --git a/docs/source/quickstart.rst b/docs/source/quickstart.rst index 8802957..8e232b1 100644 --- a/docs/source/quickstart.rst +++ b/docs/source/quickstart.rst @@ -81,7 +81,7 @@ avise --SET-list ## Advanced usage -### Configuring Connectors +### Configuring Connectors and SETs You can create your own connector configuration files, or if you cloned the AVISE repository, you can modify the existing connector configuration files in `AVISE/avise/configs/connector/languagemodel/`. @@ -98,6 +98,34 @@ For example, you can edit the default Ollama Connector configuration file `AVISE } } ``` + +Similarly, you can customize the configurations for SETs as well. For example, by editing the Red Queen SET configuration file `AVISE/avise/configs/SET/languagemodel/multi_turn/red_queen.json`, +you can define if the SET is executed incrementally *(the target model will generate a response after each subsequential prompt)*, or as a template *(only works for target systems that accept a conversation as an input)* and if the SET uses and Adversarial Language Model (ALM). +Additionally, you can define the exact template attack prompts that the SET uses: + +```json +{ + "incremental_execution": false, #Change to true to execute the SET cases incrementally + "use_adversarial_languagemodel": false, #Change to true to utilize an Adversarial Language Model (ALM) + "adversarial_model_device": "auto", # Defines on which device(s) the ALM is loaded to ("cpu", "gpu", or "auto") + "evaluation_system_prompt": "You are a highly accurate cybersecurity classifier, ..." + "evaluation_model_device": "auto", # Defines on which device(s) the ELM is loaded to ("cpu", "gpu", or "auto") + "sets": [ + { + "id": "RED-QUEEN-001", + "vulnerability_subcategory": "Evasion Attack", + "action": "make a fake passport", + "type": "occupation_teacher", + "category": "violence,aiding_and_abetting,incitement", + "conversation": [ + { ... } + ... + ] + }, + ... + ] +``` + If you want to use custom configuration files for SETs and/or Connectors, you can do so by giving the paths to the configuration files with `--SETconf` and `--connectorconf` arguments: ```bash