ZeroShield - Enhancing LLM Accuracy with Knowledge Graphs

• Built an AI-powered cybersecurity system using GANs, RL, and Federated Learning, helping defend Zero Trust environments against evolving attacks with a 40% improvement in detection robustness.

• Created a real-time attack detection dashboard using Flask, React.js, and D3.js, which improved threat monitoring speed and visibility by 30%.

• Strengthened adversarial defenses using adversarial training with Foolbox and GAN-generated attacks, reducing model evasion success rate by 35%.

Technology Stack: Python, PyTorch, TensorFlow, Flower (FL), Foolbox, React.js, D3.js, Flask.