Sultanus Salehin

Explainable AI for Medical Images (Thesis)

Building interpretable ML pipelines for accurate and transparent medical image analysis.

• Exploring how to make model decisions understandable to doctors.
• Experimenting with XAI techniques (e.g., saliency maps, Grad-CAM).
• Considering secure logging of model outputs for audit and compliance.

# Example: wrapping a model with explanation
img = load_medical_image(path)
pred, heatmap = model.predict_with_explainer(img)
store_to_audit_log(pred, heatmap_summary(heatmap))
            

Anomaly Detection in Blockchain Security

Studying how to detect malicious patterns in blockchain-based systems using ML and security indicators.

• Feature engineering on transaction graphs and wallet behaviors.
• Training ML models (Random Forest, boosting, deep models) on labeled data.
• Evaluating precision/recall tradeoffs for real-world use.

LoRa Module Research for Secure IoT

Experimental work with low-power long-range communication for secure IoT deployments.

• Measuring range, reliability, and latency under different conditions.
• Adding encryption on top of LoRa packets to defend against eavesdropping.
• Documenting configurations suitable for field deployments.

Browser & Memory Forensics Study

Academic exploration of how to extract and analyze browser and memory artifacts during incident response.

• Compared forensic tools for different browsers and OS environments.
• Outlined step-by-step workflows for evidence collection and analysis.
• Linked forensic techniques to actual attack scenarios and logs.