SecureByAI — An LLM-Powered Digital Security Assistant

🌟 Inspiration

We believe true cybersecurity is not just about defense, but also about clarity. Most people don't fall victim to scams because they lack tools — they fall because they don’t understand threats. We wanted to build something that goes beyond traditional protection — something explainable, compassionate, and intelligent. Inspired by the potential of large language models (LLMs), and by our mission to make cybersecurity accessible and mindful, we created SecureByAI.

🛡️ What it does

SecureByAI is a unified, AI-powered toolkit that helps users:

• 🧠 Ask questions about digital safety and get contextual, LLM-powered answers.
• 🔐 Analyze the strength of passwords with human-like feedback.
• 🕵️ Check if a URL might be a phishing attempt, with reasoning.
• 🤖 Use an LLM agent assistant that classifies intent and suggests tools accordingly. Instead of cold warnings or rigid rules, SecureByAI offers friendly, adaptive guidance — a digital security companion you can talk to.

🏗️ How we built it

We used:

• Next.js for the frontend (tool pages: PasswordTester, PhishingChecker, AgentAssistant).
• OpenAI's GPT-4 via API to power all security logic and explanations.
• TypeScript + Vercel Serverless Functions to handle backend API routes.
• Custom prompts and modular tool routing to future-proof agent interaction logic.
• TailwindCSS for a clean, approachable UI.
• GitHub for version control and project collaboration. Everything is built to be modular — we can easily add new tools or extend the LLM's capabilities.

🧗 Challenges we ran into

• 🧩 LLM Prompt Engineering: Getting the assistant to respond naturally and provide useful security insights required several iterations.
• 🧠 Response Hallucinations: We had to carefully design prompts and post-process outputs to avoid misleading security suggestions.
• 🕸️ Phishing detection without databases: We focused on pattern understanding and explanation using LLMs rather than relying on external URL blacklists.
• ⚙️ Tool integration: Coordinating multiple tools behind a single assistant while keeping UX smooth was complex.

🏆 Accomplishments that we're proud of

• Built a fully working LLM-driven assistant UI that connects to real tools and gives real, insightful answers.
• Created modular AI security tools (password + phishing) that respond like a human, not a scanner.
• Designed a structure that’s scalable — we can now plug in future tools like encryption helpers, privacy quizzes, and even digital wellness agents.

📚 What we learned

• LLMs are not just Q&A engines — they're UX enablers for intuitive security.
• Building for non-technical users requires empathy, not just features.
• Prompt engineering is both art and science — small wording changes can completely transform results.
• A security product is only useful if users feel safe, understood, and empowered while using it.

🚀 What's next for SecureByAI

• 🧠 Function-calling agents — enabling the assistant to invoke tools directly based on user intent.
• 🪷 Mindfulness Security Prompts — blending digital wellness with actionable advice.
• 🧩 Integration with external security platforms (password managers, identity APIs, etc).
• 🌍 Multilingual support — helping users in any language feel digitally safe.

Our ultimate goal: To build an AI security assistant that doesn't just defend — it educates, empowers, and enlightens.

Built With

• javascript
• node.js
• openai