# ai-security-bootcamp/aisb architecture

> Architecture diagram of the GitHub repository ai-security-bootcamp/aisb, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-09-30). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/ai-security-bootcamp/aisb
- Repository: https://github.com/ai-security-bootcamp/aisb

## Overview

This repository is an AI security bootcamp: learners work through practical labs covering model internals, agent control, inference defenses, training and adversarial methods, and infrastructure threats. The principal user-to-result workflow is completing a section’s exercises to build the corresponding security competency; the sections are curriculum stages, not modules in a single runtime pipeline. The README documents these capabilities and learner workflow. Sampled code illustrates individual labs, but does not establish calls between most sections, so the graph shows only documented learner-to-lab relationships.

Main entry points: [`1.1-llm-internals/section1_solution.py`](https://github.com/ai-security-bootcamp/aisb/blob/main/1.1-llm-internals/section1_solution.py) (LLM Internals), [`1.2-logprobs/section2_solution.py`](https://github.com/ai-security-bootcamp/aisb/blob/main/1.2-logprobs/section2_solution.py) (Log Probabilities), [`1.3-instruction-hierarchies/section3_solution.py`](https://github.com/ai-security-bootcamp/aisb/blob/main/1.3-instruction-hierarchies/section3_solution.py) (Instruction Hierarchies), [`1.4-prompt-injection/section4_solution.py`](https://github.com/ai-security-bootcamp/aisb/blob/main/1.4-prompt-injection/section4_solution.py) (Prompt Injection)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_internals["LLM Internals"]
  node_llm_internals["LLM Internals"]
  node_logprobs["Log Probabilities"]
  node_instruction_hierarchies["Instruction Hierarchies"]
  node_prompt_injection["Prompt Injection"]
end

subgraph group_control["Agents And Control"]
  node_coding_agents["Coding Agents"]
  node_monitoring["AI-Control Monitoring"]
  node_control_protocols["Control Protocols"]
  node_safety_simulator["Safety Simulator"]
end

subgraph group_inference["Inference Security"]
  node_tokenization["Tokenization"]
  node_jailbreaking["Jailbreaking"]
  node_guardrails["Guardrails"]
  node_distillation["Knowledge Distillation"]
  node_weight_extraction["Weight Extraction"]
end

subgraph group_model_security["Model Security"]
  node_model_editing["Model Editing"]
  node_backdooring["Backdooring"]
  node_undo_safety["Undoing Safety Tuning"]
  node_safety_finetuning["Safety Fine-Tuning"]
  node_adversarial_vision["Adversarial Vision"]
  node_adversarial_language["Adversarial Language"]
  node_prefix_tuning["Prefix Tuning"]
  node_watermarking["Image Watermarking"]
end

subgraph group_infrastructure["Infrastructure Security"]
  node_side_channel["Side-Channel Monitoring"]
  node_rand_report["Model Weight Security"]
  node_threat_modeling["Threat Modeling"]
  node_container_toolkit["Container Toolkit Vulnerability"]
  node_gpu_rowhammer["GPU RowHammer"]
end

node_learner(("Bootcamp learner"))

node_learner -->|"completes lab"| node_llm_internals
node_learner -->|"completes lab"| node_logprobs
node_learner -->|"completes lab"| node_instruction_hierarchies
node_learner -->|"completes lab"| node_prompt_injection
node_learner -->|"completes lab"| node_coding_agents
node_learner -->|"completes lab"| node_monitoring
node_learner -->|"completes lab"| node_control_protocols
node_learner -->|"completes lab"| node_safety_simulator
node_learner -->|"completes lab"| node_tokenization
node_learner -->|"completes lab"| node_jailbreaking
node_learner -->|"completes lab"| node_guardrails
node_learner -->|"completes lab"| node_distillation
node_learner -->|"completes lab"| node_weight_extraction
node_learner -->|"completes lab"| node_model_editing
node_learner -->|"completes lab"| node_backdooring
node_learner -->|"completes lab"| node_undo_safety
node_learner -->|"completes lab"| node_safety_finetuning
node_learner -->|"completes lab"| node_adversarial_vision
node_learner -->|"completes lab"| node_adversarial_language
node_learner -->|"completes lab"| node_prefix_tuning
node_learner -->|"completes lab"| node_watermarking
node_learner -->|"completes lab"| node_side_channel
node_learner -->|"completes lab"| node_rand_report
node_learner -->|"completes lab"| node_threat_modeling
node_learner -->|"completes lab"| node_container_toolkit
node_learner -->|"completes lab"| node_gpu_rowhammer

click node_llm_internals "https://github.com/ai-security-bootcamp/aisb/blob/main/1.1-llm-internals/section1_solution.py"
click node_logprobs "https://github.com/ai-security-bootcamp/aisb/blob/main/1.2-logprobs/section2_solution.py"
click node_instruction_hierarchies "https://github.com/ai-security-bootcamp/aisb/blob/main/1.3-instruction-hierarchies/section3_solution.py"
click node_prompt_injection "https://github.com/ai-security-bootcamp/aisb/blob/main/1.4-prompt-injection/section4_solution.py"
click node_coding_agents "https://github.com/ai-security-bootcamp/aisb/blob/main/2.1-coding-agents/section1_solution.py"
click node_monitoring "https://github.com/ai-security-bootcamp/aisb/blob/main/2.2-monitoring/section2_solution.py"
click node_control_protocols "https://github.com/ai-security-bootcamp/aisb/blob/main/2.3-control-protocols/section3_solution.py"
click node_safety_simulator "https://github.com/ai-security-bootcamp/aisb/blob/main/2.4-safety-simulator/section4_solution.py"
click node_tokenization "https://github.com/ai-security-bootcamp/aisb/blob/main/3.1-tokenization/section1_solution.py"
click node_jailbreaking "https://github.com/ai-security-bootcamp/aisb/blob/main/3.2-jailbreaking/section2_solution.py"
click node_guardrails "https://github.com/ai-security-bootcamp/aisb/blob/main/3.3-guardrails/section3_solution.py"
click node_distillation "https://github.com/ai-security-bootcamp/aisb/blob/main/3.4-knowledge-distillation/section4_solution.py"
click node_weight_extraction "https://github.com/ai-security-bootcamp/aisb/blob/main/3.5-weight-extraction/section5_solution.py"
click node_model_editing "https://github.com/ai-security-bootcamp/aisb/blob/main/4.1-model-editing/section1_solution.py"
click node_backdooring "https://github.com/ai-security-bootcamp/aisb/blob/main/4.2-backdooring/section2_solution.py"
click node_undo_safety "https://github.com/ai-security-bootcamp/aisb/blob/main/4.3-undoing-safety-finetuning/section3_solution.py"
click node_safety_finetuning "https://github.com/ai-security-bootcamp/aisb/blob/main/4.4-safety-finetuning/section4_solution.py"
click node_adversarial_vision "https://github.com/ai-security-bootcamp/aisb/blob/main/5.1-adversarial-vision/section1_solution.py"
click node_adversarial_language "https://github.com/ai-security-bootcamp/aisb/blob/main/5.2-adversarial-language/section2_solution.py"
click node_prefix_tuning "https://github.com/ai-security-bootcamp/aisb/blob/main/5.3-prefix-tuning/section3_solution.py"
click node_watermarking "https://github.com/ai-security-bootcamp/aisb/blob/main/5.4-watermarking/section4_solution.py"
click node_side_channel "https://github.com/ai-security-bootcamp/aisb/blob/main/6.1-side-channel-monitoring/section1_solution.py"
click node_rand_report "https://github.com/ai-security-bootcamp/aisb/blob/main/7.1-rand-report/section1_solution.py"
click node_threat_modeling "https://github.com/ai-security-bootcamp/aisb/blob/main/7.2-threat-modeling/section2_solution.py"
click node_container_toolkit "https://github.com/ai-security-bootcamp/aisb/blob/main/7.3-nvidia-container-toolkit/section3_solution.py"
click node_gpu_rowhammer "https://github.com/ai-security-bootcamp/aisb/blob/main/7.4-gpu-rowhammer/section4_solution.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_llm_internals,node_logprobs,node_instruction_hierarchies,node_prompt_injection toneBlue
class node_coding_agents,node_monitoring,node_control_protocols,node_safety_simulator toneAmber
class node_tokenization,node_jailbreaking,node_guardrails,node_distillation,node_weight_extraction toneMint
class node_model_editing,node_backdooring,node_undo_safety,node_safety_finetuning,node_adversarial_vision,node_adversarial_language,node_prefix_tuning,node_watermarking toneRose
class node_side_channel,node_rand_report,node_threat_modeling,node_container_toolkit,node_gpu_rowhammer,node_learner toneIndigo
```

## Components

### LLM Internals

- [LLM Internals](https://github.com/ai-security-bootcamp/aisb/blob/main/1.1-llm-internals/section1_solution.py): `1.1-llm-internals/section1_solution.py`
- [Log Probabilities](https://github.com/ai-security-bootcamp/aisb/blob/main/1.2-logprobs/section2_solution.py): `1.2-logprobs/section2_solution.py`
- [Instruction Hierarchies](https://github.com/ai-security-bootcamp/aisb/blob/main/1.3-instruction-hierarchies/section3_solution.py): `1.3-instruction-hierarchies/section3_solution.py`
- [Prompt Injection](https://github.com/ai-security-bootcamp/aisb/blob/main/1.4-prompt-injection/section4_solution.py): `1.4-prompt-injection/section4_solution.py`

### Agents And Control

- [Coding Agents](https://github.com/ai-security-bootcamp/aisb/blob/main/2.1-coding-agents/section1_solution.py): `2.1-coding-agents/section1_solution.py`
- [AI-Control Monitoring](https://github.com/ai-security-bootcamp/aisb/blob/main/2.2-monitoring/section2_solution.py): `2.2-monitoring/section2_solution.py`
- [Control Protocols](https://github.com/ai-security-bootcamp/aisb/blob/main/2.3-control-protocols/section3_solution.py): `2.3-control-protocols/section3_solution.py`
- [Safety Simulator](https://github.com/ai-security-bootcamp/aisb/blob/main/2.4-safety-simulator/section4_solution.py): `2.4-safety-simulator/section4_solution.py`

### Inference Security

- [Tokenization](https://github.com/ai-security-bootcamp/aisb/blob/main/3.1-tokenization/section1_solution.py): `3.1-tokenization/section1_solution.py`
- [Jailbreaking](https://github.com/ai-security-bootcamp/aisb/blob/main/3.2-jailbreaking/section2_solution.py): `3.2-jailbreaking/section2_solution.py`
- [Guardrails](https://github.com/ai-security-bootcamp/aisb/blob/main/3.3-guardrails/section3_solution.py): `3.3-guardrails/section3_solution.py`
- [Knowledge Distillation](https://github.com/ai-security-bootcamp/aisb/blob/main/3.4-knowledge-distillation/section4_solution.py): `3.4-knowledge-distillation/section4_solution.py`
- [Weight Extraction](https://github.com/ai-security-bootcamp/aisb/blob/main/3.5-weight-extraction/section5_solution.py): `3.5-weight-extraction/section5_solution.py`

### Model Security

- [Model Editing](https://github.com/ai-security-bootcamp/aisb/blob/main/4.1-model-editing/section1_solution.py): `4.1-model-editing/section1_solution.py`
- [Backdooring](https://github.com/ai-security-bootcamp/aisb/blob/main/4.2-backdooring/section2_solution.py): `4.2-backdooring/section2_solution.py`
- [Undoing Safety Tuning](https://github.com/ai-security-bootcamp/aisb/blob/main/4.3-undoing-safety-finetuning/section3_solution.py): `4.3-undoing-safety-finetuning/section3_solution.py`
- [Safety Fine-Tuning](https://github.com/ai-security-bootcamp/aisb/blob/main/4.4-safety-finetuning/section4_solution.py): `4.4-safety-finetuning/section4_solution.py`
- [Adversarial Vision](https://github.com/ai-security-bootcamp/aisb/blob/main/5.1-adversarial-vision/section1_solution.py): `5.1-adversarial-vision/section1_solution.py`
- [Adversarial Language](https://github.com/ai-security-bootcamp/aisb/blob/main/5.2-adversarial-language/section2_solution.py): `5.2-adversarial-language/section2_solution.py`
- [Prefix Tuning](https://github.com/ai-security-bootcamp/aisb/blob/main/5.3-prefix-tuning/section3_solution.py): `5.3-prefix-tuning/section3_solution.py`
- [Image Watermarking](https://github.com/ai-security-bootcamp/aisb/blob/main/5.4-watermarking/section4_solution.py): `5.4-watermarking/section4_solution.py`

### Infrastructure Security

- [Side-Channel Monitoring](https://github.com/ai-security-bootcamp/aisb/blob/main/6.1-side-channel-monitoring/section1_solution.py): `6.1-side-channel-monitoring/section1_solution.py`
- [Model Weight Security](https://github.com/ai-security-bootcamp/aisb/blob/main/7.1-rand-report/section1_solution.py): `7.1-rand-report/section1_solution.py`
- [Threat Modeling](https://github.com/ai-security-bootcamp/aisb/blob/main/7.2-threat-modeling/section2_solution.py): `7.2-threat-modeling/section2_solution.py`
- [Container Toolkit Vulnerability](https://github.com/ai-security-bootcamp/aisb/blob/main/7.3-nvidia-container-toolkit/section3_solution.py): `7.3-nvidia-container-toolkit/section3_solution.py`
- [GPU RowHammer](https://github.com/ai-security-bootcamp/aisb/blob/main/7.4-gpu-rowhammer/section4_solution.py): `7.4-gpu-rowhammer/section4_solution.py`

### Other components

- **Bootcamp learner**

## Connections

- Bootcamp learner → LLM Internals: completes lab
- Bootcamp learner → Log Probabilities: completes lab
- Bootcamp learner → Instruction Hierarchies: completes lab
- Bootcamp learner → Prompt Injection: completes lab
- Bootcamp learner → Coding Agents: completes lab
- Bootcamp learner → AI-Control Monitoring: completes lab
- Bootcamp learner → Control Protocols: completes lab
- Bootcamp learner → Safety Simulator: completes lab
- Bootcamp learner → Tokenization: completes lab
- Bootcamp learner → Jailbreaking: completes lab
- Bootcamp learner → Guardrails: completes lab
- Bootcamp learner → Knowledge Distillation: completes lab
- Bootcamp learner → Weight Extraction: completes lab
- Bootcamp learner → Model Editing: completes lab
- Bootcamp learner → Backdooring: completes lab
- Bootcamp learner → Undoing Safety Tuning: completes lab
- Bootcamp learner → Safety Fine-Tuning: completes lab
- Bootcamp learner → Adversarial Vision: completes lab
- Bootcamp learner → Adversarial Language: completes lab
- Bootcamp learner → Prefix Tuning: completes lab
- Bootcamp learner → Image Watermarking: completes lab
- Bootcamp learner → Side-Channel Monitoring: completes lab
- Bootcamp learner → Model Weight Security: completes lab
- Bootcamp learner → Threat Modeling: completes lab
- Bootcamp learner → Container Toolkit Vulnerability: completes lab
- Bootcamp learner → GPU RowHammer: completes lab