No Phish AI

Inspiration

Though many security tools presently exist, cyber attacks are continuing to rise with phishing among the most common attack vectors  While certain aspects of phishing can be well controlled with current technologies, there is a need to further advance our defenses And with recent advancements in LLM, we now have additional opportunities to approach this differently This is where the “No Phish AI” application which provides an LLM Driven Security Layer to defend against the rising complexity of phishing attacks

What it does

Analyze and detect phishing pages with greater precision with LLM Live Application : https://no-phish-ai-irudykcscvemhc4enmtutd.streamlit.app/

How we built it

GPT-4 , LangChain,Langsmith, MongoDB,Streamlit

Challenges we ran into

Sometimes inconsistent security scoring

Accomplishments that we're proud of

Innovative Security Layer for defending from Phishing Attacks

What we learned

LLM's Classification responses are different for different period runs.

What's next for No Phish AI

Expanding Results Databases Incorporating Advanced Retrieval Augmented Generation (RAG) Continuing to Refine the Model for Robustness

Built With

• gpt-4
• langchain
• langsmith
• mongodb
• streamlit