cyber-security-resources/ai_research/README.md

# AI Security Research Resources

## Langchain Resources
- [Tools, Learning, Notebooks, Bots, Agent examples, etc.](https://github.com/The-Art-of-Hacking/h4cker/blob/master/ai_research/LangChain/README.md)

## LLM Frameworks
- [Popular LLM Frameworks](https://github.com/The-Art-of-Hacking/h4cker/tree/master/ai_research/LLM-frameworks)
  
## AI Security Best Practices and Tools
- [High-Level AI Security Best Practices](<AI Security Best Practices/AI-security-tools-and-frameworks.md>)
- [Homomorphic-Encryption](<AI Security Best Practices/homomorphic-encryption.md>)
- [AI Security Tools and Frameworks](<AI Security Best Practices/AI-security-tools-and-frameworks.md>)
- [AI Secure Deployment Tips](<AI Security Best Practices/secure-deployment.md>)
- [AI Secure Design Tips](<AI Security Best Practices/secure-design.md>)
- [Threat Modeling](<AI Security Best Practices/threat-modeling.md>)

## AI Security Resources from Omar's Training Sessions
- [Cybersecurity Learning Prompts](https://github.com/santosomar/chatgpt-cybersecurity-prompts)
- [Networking Prompts](https://github.com/santosomar/chatgpt-networking-prompts)
- [Programming Learning Prompts](https://github.com/santosomar/chatgpt-programming-prompts)

## AI Ethics and Privacy Resources
- [AI Ethics and Privacy Resources](https://github.com/The-Art-of-Hacking/h4cker/tree/master/ai_research/ethics_privacy)

## Tools & Methods for Data Collection, Cleaning, and Analysis:

- **Data Collection**:
  - APIs and SDKs
  - Wireless transmission

### Data Cleaning:

3. **Pandas**:
   - **Example**: Cleaning a dataset with missing values using Pandas before training a machine learning model.
   - **Relevant Link**: [Pandas Documentation](https://pandas.pydata.org/pandas-docs/stable/index.html)
   - **Usage in HAR and AI**: Pandas can be used to structure and clean sensor data, making it suitable for training AI models capable of recognizing complex patterns in human activity data.

4. **Sci-kit learn**:
   - **Example**: Using Sci-kit learn for feature selection and removing irrelevant features from a dataset.
   - **Relevant Link**: [Sci-kit learn Documentation](https://scikit-learn.org/stable/)
   - **Usage in HAR and AI**: Sci-kit learn offers various tools for data preprocessing, which is a vital step in preparing data for AI algorithms, enhancing the performance of the models in HAR applications.

### Data Analysis:

5. **TensorFlow**:
   - **Example**: Developing a deep learning model using TensorFlow to classify different activities based on sensor data.
   - **Relevant Link**: [TensorFlow Documentation](https://www.tensorflow.org/learn)
   - **Usage in HAR and AI**: TensorFlow provides a comprehensive platform for developing and training AI models capable of analyzing and recognizing patterns in human activity data.

6. **Keras**:
   - **Example**: Using Keras to create a convolutional neural network (CNN) for image recognition, an essential task in AI.
   - **Relevant Link**: [Keras Documentation](https://keras.io/getting_started/intro_to_keras_for_engineers/)
   - **Usage in HAR and AI**: Keras simplifies the process of building and optimizing neural networks, a crucial component in AI, to analyze human activity data more effectively and make predictions.

### Visualization and Further Analysis:

7. **Matplotlib**:
   - **Example**: Using Matplotlib to visualize the distribution of different activities within a dataset.
   - **Relevant Link**: [Matplotlib Documentation](https://matplotlib.org/stable/contents.html)
   - **Usage in HAR and AI**: Visualization of data is essential in AI to understand underlying patterns and trends in data, aiding in the better development and tuning of models for HAR.

8. **Seaborn**:
   - **Example**: Creating a heatmap using Seaborn to visualize the correlation between different features in a dataset.
   - **Relevant Link**: [Seaborn Documentation](https://seaborn.pydata.org/)
   - **Usage in HAR and AI**: Seaborn can enhance data visualization in AI, assisting in identifying relationships and patterns in data which can influence the development and performance of HAR models.