Fr.
Conceicao Rodrigues College of Engineering
Father Agnel Ashram, Bandstand, Bandra - west, Mumbai-50
Department of Computer Engineering
PROJECT PROPOSAL
Group Information:
Roll No Student Name Div A/ B
101 (Group Leader)
102
103
104
TITLE OF THE PROJECT:
AI-Powered Adaptive Threat Detection and Secure Traffic Analytics System
BROAD AREA:
Artificial Intelligence and Machine Learning (AI/ML); Cybersecurity; Network Security and
Analytics
Category of the Project:
Research based
ABSTRACT (maximum 300 words):
This project proposes an AI-powered cybersecurity framework for detecting network intrusions,
analyzing malicious traffic, generating adaptive threat scores, recommending security
responses, and presenting explainable AI-driven insights. Traditional intrusion detection
systems often stop at raising alerts, which leaves security teams with limited guidance on how
serious a threat is and what should happen next. The proposed system closes that gap by
combining machine learning-based traffic classification, adaptive threat scoring, dynamic
encryption recommendations, secure traffic routing guidance, and explainable analytics in a
single platform. The workflow uses network traffic capture, feature extraction, model inference,
and decision modules that convert raw packet data into actionable security intelligence. The
expected outcome is a unified dashboard that helps administrators understand attack severity,
interpret why a packet or flow was flagged, and choose safer routing and encryption
responses. The project is designed as a practical research prototype for modern enterprise
and campus networks.
MOTIVATION (maximum 100 words):
Modern networks generate huge volumes of traffic, and manual monitoring is too slow for
today's attack patterns. A system that not only detects suspicious activity but also explains the
threat and suggests a response can reduce delay, confusion, and overreaction during
incidents.
PROBLEM STATEMENT:
Page 1
Existing intrusion detection systems mainly focus on attack detection and lack adaptive,
explainable, and intelligent response mechanisms for modern cybersecurity threats.
SDGs Mapped:
SDG 9 - Industry, Innovation and Infrastructure; SDG 11 - Sustainable Cities and
Communities; SDG 16 - Peace, Justice and Strong Institutions; SDG 17 - Partnerships for the
Goals; SDG 8 - Decent Work and Economic Growth
OBJECTIVES:
• Detect malicious network traffic using machine learning.
• Generate adaptive threat scores based on traffic behavior and risk indicators.
• Provide explainable AI-based threat reasoning for flagged packets and flows.
• Recommend dynamic encryption levels according to threat severity.
• Suggest secure traffic routing for suspicious or high-risk traffic.
• Visualize cybersecurity analytics through a real-time dashboard.
METHODOLOGY (TENTATIVE IF ANY):
• Collect traffic using Wireshark, Tshark, Scapy, and benchmark datasets.
• Extract relevant packet and flow features and clean the data for modeling.
• Train intrusion detection models using Random Forest and XGBoost.
• Compute adaptive threat scores and classify threat severity.
• Generate explainability outputs using SHAP and LIME.
• Recommend encryption and routing actions based on the threat score.
• Display results through a Streamlit dashboard backed by FastAPI and Python.
Page 2
HARDWARE / SOFTWARE REQUIREMENTS:
Hardware: Standard laptop or workstation for development and testing.
Software: Python, Streamlit, FastAPI, Scikit-Learn, XGBoost, Pandas, NumPy, SHAP, LIME,
Wireshark, Tshark, Scapy, MongoDB, Plotly, Matplotlib, PyCryptodome, Cryptography.
INNOVATIVENESS:
• Adaptive threat scoring instead of simple binary alerting.
• Explainable AI-based security decisions for better trust and auditability.
• Dynamic encryption adaptation based on detected risk.
• Intelligent secure traffic routing recommendations.
• Unified cybersecurity analytics platform combining detection, explanation, and response
guidance.
SOCIETAL RELEVANCE:
The project supports safer digital infrastructure for organizations that rely on constant network
communication. It is relevant to healthcare, education, smart cities, public institutions, and
enterprise systems where downtime or compromise can affect large numbers of users. By
improving detection speed, response quality, and interpretability, the system can help reduce
security incidents and strengthen trust in connected services.
EXPECTED OUTPUTS:
• Intrusion detection results and threat severity scores.
• Threat heatmaps and security analytics dashboard.
• Explainable AI reports for suspicious traffic.
• Encryption and routing recommendations.
RESEARCH CONTRIBUTION:
The project extends traditional intrusion detection systems by integrating adaptive threat
scoring, explainable AI, dynamic encryption adaptation, and intelligent traffic routing
recommendations within a unified cybersecurity framework.
FUTURE SCOPE:
• Deep learning-based threat detection.
• Real-time enterprise deployment.
• Zero Trust security integration.
• Cloud security analytics.
• Federated learning for distributed cybersecurity.
Page 3