Research by our students

A practical password-strength analyzer that combines basic security checks with a lightweight machine-learning model trained to recognise patterns associated with password guessability, producing an understandable strength category and feedback.

An automated system that estimates whether a news item shares patterns associated with misinformation, producing a probabilistic confidence estimate rather than claiming to determine truth, and weighing satire, bias and evolving claims.

Examines remote electronic voting and its central problem — proving a vote was counted without revealing it — reviewing homomorphic encryption, mix-nets and zero-knowledge proofs, and real-world experience such as Estonia's system.

A Python toolkit combining evidence acquisition, cryptographic hashing, metadata analysis, deleted-file identification, timelines and chain-of-custody records — with attention to the risks forensic automation introduces.

A digital insurance platform for policy administration and claims, with machine learning used to flag potentially fraudulent claims for human review, grounded in published fraud-detection research.

A personal finance system recording transactions manually and from bank/UPI SMS, with budgets, reports, expense prediction and unusual-spending detection drawn from financial anomaly-detection research.

A design-stage early-warning architecture: heart-rate, SpO₂ and temperature sensors on a microcontroller, local threshold checks, a cloud dashboard and automated SMS/email alerts for remote monitoring.

A web application mapping reported symptoms to the most likely condition with a trained classifier — positioned as a preliminary, supportive tool to help decide whether to see a doctor, never as a diagnosis.

A web-based finance coach that analyses income, spending, savings and goals, scores financial health from indicators such as savings rate and expense-to-income ratio, and gives personalised budgeting guidance.

A lightweight early-warning system that plants canary (honey) files in monitored folders and treats any access, change, rename or deletion of them as a strong signal of ransomware activity.

A lightweight, explainable change-detection system: two satellite images of the same area are aligned and differenced, then thresholding and morphological processing isolate real surface change.