Solo capstone project: a framework for generating and detecting SMS phishing (smishing) attacks. Built a Word-CNN spam classifier (Keras/TensorFlow) trained on the SMS Spam Collection dataset and deployed via Flask, plus a rule-based scoring engine that scores messages 0-31 based on indicators such as suspicious URLs, urgent language, unusual formatting, and misspellings. An early version of the detection pipeline later extended in the MSc dissertation below.
My solo BSc capstone at Ashesi University (2023): an open-source framework for generating and detecting SMS phishing (smishing) attacks, built to help researchers and developers test defences against this attack vector.
This project is the direct predecessor to my MSc dissertation. Both follow the same shape — ingest a message, extract features, classify with ML, and produce a risk signal — but applied to SMS in the capstone and email in the dissertation, with the dissertation adding transformer models (DistilBERT, BERT-base) on top of the classical/CNN approach used here.