Completed & Published
🏆 Journal Publication
🎙 Conference Paper
Opaline Attachment Defence System
An AI-powered email security system that proactively detects and neutralizes malicious email attachments in real time. Instead of relying on traditional signature-based scanning, the system uses a DistilBERT-based CANet model to classify attachments as safe or malicious — then converts flagged files into static PNG images, stripping all executable content before the user interacts with them.
Performance vs Traditional Antivirus
94%
Detection
Accuracy
+22% vs baseline
87%
Zero-Day
Detection
+56% vs baseline
6s
Processing
Speed
50% faster
13%
Attacker
Evasion
−56% vs baseline
6.2
Impact
Factor
INDJCST Journal
DistilBERT-based CANet model classifies email attachments using NLP preprocessing, TF-IDF feature extraction, and word embeddings — detects both known and zero-day threats
Sandboxed converter transforms malicious PDFs, DOCX, and TXT files into static PNG images — neutralizing all executable content before user interaction
Real-time IMAP inbox monitor with automated threat detection, attachment deletion, and instant email alert dispatch via Flask-Mail and SMTP
Full-stack system: Python Flask backend, React JS frontend, MySQL database with 8 normalized tables — users, predictions, alerts, and conversion logs
System Pipeline
Email Inbox→
Extract Attachment→
NLP Preprocessing→
DistilBERT CANet→
PNG Sandbox→
Alert User
📚 INDJCST e-ISSN: 2583-5300 · Vol.5, Issue 1, PP 350–354 · Impact Factor: 6.211 · UGC-CARE Listed
🎙 ICNTSETM’26 3rd International Conference · Hybrid Mode · March 2026
Python FlaskReact JS
DistilBERTNLP / TF-IDFMySQL
TensorFlowCANet Model
IMAP / SMTPCybersecurityMalware Detection
