Home About IT Training Consultancy Research Scholarship Gallery Contact
Research to Publication

Arrive with an interest. Leave with a paper under review.

Six-month mentored tracks in AI and cybersecurity. Supervision is weekly, methodology is challenged hard before anything is submitted, and support continues through the revision cycle if the first decision comes back negative.

96
Papers published
4
Active tracks
6 months
Track duration
~33%
Accepted first attempt
How it works

Structured for the first six weeks. Yours after that.

The hardest part of a first paper is not the model — it is finding a question worth answering and defending the method. That is what the supervision is for.

Rejection is planned for
Supervision continues through revision and resubmission. You are not dropped when the first decision arrives.
1

Weeks 1–6 · Literature and gap

Systematic mapping of the field, then a defensible research question grounded in an actual gap rather than a hunch.

2

Weeks 7–14 · Method and data

Dataset construction or selection, experimental design, and a reproducible pipeline reviewed before you run anything at scale.

3

Weeks 15–20 · Experiments

Implementation, evaluation against honest published baselines, and ablation. The implementation clinic runs weekly.

4

Weeks 21–26 · Writing and submission

Manuscript drafting, internal hostile review, venue selection and submission.

Tracks

Choose a research direction

Places are limited per track and selected on fit. Each track is supervised by the named researcher.

Applied ML 8 places

ML-Based Threat Detection

You arrive with an interest and leave with a paper under review. The first six weeks are structured: literature mapping, gap identification, and a defensible research question. The remainder is your study — dataset construction, model development, evaluation against honest baselines, and writing. Supervision is weekly and the methodology is challenged hard before anything is submitted.

6 months 10–12 hrs/week
Example questions
Concept drift in malware family classification Few-shot detection of novel attack techniques Explainability for SOC analyst trust
  • A defensible research question grounded in a literature gap
  • A reproducible experimental pipeline
  • Evaluation against honest, published baselines
  • A manuscript submitted to a peer-reviewed venue

Prerequisites: Python and basic machine learning. A bachelor's degree in a computing discipline, or equivalent practical work.

৳45,000
Supervised by Dr. Karim Hassan
Adversarial ML 6 places

Adversarial AI & Model Robustness

Security models are deployed against opponents who adapt. This track studies evasion and poisoning attacks against detection systems, then the defences — adversarial training, certified robustness, ensemble strategies — and honestly evaluates which of them hold up outside the paper they were published in.

6 months 10–12 hrs/week
Example questions
Adaptive evasion against ensemble malware detectors Poisoning resilience in federated threat intelligence Certified robustness bounds for network IDS
  • Implement evasion and poisoning attacks against real models
  • Evaluate defensive techniques under adaptive attack
  • Understand the robustness–accuracy trade-off empirically
  • A manuscript submitted to a peer-reviewed venue

Prerequisites: Solid Python and PyTorch. Prior ML coursework or project experience.

৳45,000
Supervised by Dr. Elena Russo
Network Security 8 places

Network Behaviour Analytics

When traffic is encrypted, detection has to work from behaviour: timing, volume, graph structure, sequence. This track studies what can genuinely be inferred, where the published claims overstate themselves, and how to build detection that works on real enterprise traffic rather than a clean academic dataset.

6 months 8–10 hrs/week
Example questions
Graph-based lateral movement detection in AD environments Insider threat signals from access sequence modelling Encrypted traffic classification without payload inspection
  • Build graph and sequence models over network telemetry
  • Construct and validate a realistic evaluation dataset
  • Quantify detection performance honestly under class imbalance
  • A manuscript submitted to a peer-reviewed venue

Prerequisites: Networking fundamentals and Python. Familiarity with packet analysis helps.

৳42,000
Supervised by Dr. Amina Khan
Applied ML 6 places

AI for Vulnerability Discovery

Covers learned vulnerability detection over source code and binaries, guided fuzzing, and exploitability prediction for patch prioritisation. A practical track — participants typically end up with both a paper and a tool that finds something real.

6 months 10–12 hrs/week
Example questions
Transformer models for vulnerable code detection ML-guided seed selection in coverage fuzzing Exploitability prediction from static features
  • Apply learned models to code and binary analysis
  • Build ML-guided fuzzing harnesses
  • Model exploitability for patch prioritisation
  • A manuscript submitted to a peer-reviewed venue

Prerequisites: Python, plus C or assembly reading ability. Some program analysis background is an advantage.

৳45,000
Supervised by Md. Kamal Uddin

Final-year and postgraduate students

You have the coursework but no route to a first publication. This is the most common profile on the tracks.

Working professionals

You want to move into research or strengthen a postgraduate application. Evening and weekend supervision is normal.

Faculty and lab supervisors

Several universities send postgraduate students through the tracks and keep the methodology discipline afterwards.

Supervision

The researchers who supervise your work

You are supervised by a named person, weekly, for the full six months.

ML CTO
Maria Lopez
Chief Technology Officer
CKS, AWS Solutions Architect Professional
20 years · Cyber range architecture
View profile
KH Research Director
Dr. Karim Hassan
Research Director — ML Threat Detection
IEEE Member
14 years · Deep learning for security
View profile
AZ Senior Researcher
Dr. Amina Khan
Senior Researcher — Network Behaviour Analytics
IEEE Member
12 years · Network behaviour analytics
View profile
Collaborations

Universities and institutions we publish with

ID ICT Division Programme partner
BU BUET Research collaboration
NS North South University Research collaboration
IE IEEE Bangladesh Publication partner
Participant feedback

What researchers said afterwards

Two papers submitted and one accepted at an IEEE workshop during the six-month track. The mentorship on methodology was the part I genuinely could not have managed alone.

NT Nusrat Tabassum MSc Researcher

I sent three of my postgraduate students through the behaviour analytics track. All three came out with submissions, and the methodology discipline carried back into my own lab.

IH Dr. Imran Hossain Assistant Professor
Limited places

Applications are read by the supervising researcher.

A rough idea is enough to apply. We will tell you within five working days whether the fit is right and which track suits the question.