Hybrid Trust Model for Cloud Service Provider (CSP) Selection

A bidirectional trust model integrating NLP-based requirement extraction and malicious feedback filtering for reliable CSP recommendations.

📝 Overview

This project proposes a hybrid trust model to address the challenges of Cloud Service Provider (CSP) selection by:

1. Bidirectional Trust Evaluation: Combines CSP trust scores (based on QoS parameters) and user feedback trust (collaborative filtering).
2. NLP-Powered Requirement Extraction: Allows users to input requirements via text or numerical weights using a rule-based NLP module (98% accuracy).
3. Malicious Feedback Filtering: Detects fake reviews using multivariate outlier detection (IQR method) without requiring frequent feedback updates.

Key Result: Outperforms existing models by 59.01% (vs. QoS-only) and 17.68% (vs. feedback-only).

---

🛠️ Installation

1. Clone the repository:

   git clone https://github.com/BhanuPrakashNani/hybrid-trust-model.git  
   cd hybrid-trust-model  

2. Install dependencies:

   pip install -r requirements.txt  # Includes spaCy, pandas, scikit-learn, etc.  
   python -m spacy download en_core_web_sm  # NLP model  

---

🚀 Usage

1. Trust Calculation

Run the hybrid trust model:

python src/trust_calculation.py \  
    --user_requirements "uptime: high, downtime: very low" \  
    --num_recommendations 5  

Input formats: Text (e.g., "uptime: high") or numerical weights (e.g., {"uptime": 4.5}).

2. Malicious Feedback Detection

Filter outliers from feedback data:

python src/feedback_filter.py --feedback_csv data/feedback.csv  

3. NLP Weight Extractor

Test the rule-based NLP module:

python src/nlp_extractor.py --text "I need very high uptime and low downtime"  

Output: {"uptime": 5, "downtime": 1}

---

📂 Repository Structure

├── data/                    # Sample datasets  
│   ├── csp_parameters.csv   # CSP QoS promises  
│   └── feedback.csv         # User feedback logs  
├── src/  
│   ├── trust_calculation.py # Hybrid trust algorithm  
│   ├── feedback_filter.py   # Outlier detection  
│   └── nlp_extractor.py     # NLP weight extraction  
├── docs/  
│   └── ICCCNT_2021_Paper.pdf # Published paper  
└── requirements.txt         # Dependencies  

---

📊 Results

• Optimal Weights: 0.3 (QoS trust) + 0.7 (feedback trust).
• Accuracy: 70.2% successful recommendations (vs. 26.4% for QoS-only).
• Malicious Feedback: 0.3% improvement in recommendations after filtering.

---

📜 Citation

If you use this work, please cite:

@inproceedings{poluparthi2021hybrid,  
  title={A Hybrid Trust Model for Cloud Service Provider Selection with NLP Support and Malicious User Feedback Filtering},  
  author={Poluparthi, Bhanu Prakash and Kishan, G. Mani and Praneeth, V. Bala Sai and Manikanta, A. and Sarath, Greeshma},  
  booktitle={2021 12th International Conference on Computing Communication and Networking Technologies (ICCCNT)},  
  year={2021},  
  organization={IEEE}  
}  

---

🤝 Contributing

Contributions are welcome! Open an issue or submit a PR for:

• Expanding QoS parameters.
• Enhancing NLP rule-based models.
• Optimizing outlier detection.

---

📄 License

MIT © Bhanu Prakash Poluparthi

---