NeuroShield IDS: AI-Powered Intrusion and Scam Detection System

Abstract

Abstract Network intrusions and phishing attacks are growing problems for digital infrastructure. Spam messages and fraudulent legal notices add to the threat. Most Intrusion Detection Systems and spam filters still work separately, though. They also rely on static, rule-based logic. That does not hold up well against adaptive adversaries and new attack patterns. This paper presents NeuroShield IDS, an AI-powered system. It brings network intrusion detection and NLP-based scam/phishing classification into one place. Random Forest handles network-threat classification on the NSL-KDD and CICIDS2017 datasets. Naive Bayes with TF IDF vectorization handles scam and phishing detection on the SMS Spam Collection Dataset. A unified risk-scoring engine combines the severity signals from both modules. It sends real-time alerts through a Flask/Streamlit dashboard. Our experiments show over 97% accuracy for network intrusion detection. Scam classification stays above 95%. F1-scores hold up strong across every threat category. Keywords—Intrusion Detection System (IDS); Machine Learning; Random Forest; Naive Bayes; TF-IDF; Natural Language Processing (NLP); Phishing Detection; Cybersecurity; NSL-KDD; CICIDS2017; Risk Scoring