DYNAS presented at NSS 2026
My first-authored paper on dynamic, IDS-aware network penetration testing was accepted on 20 October and presented on 1 November 2026 at the 20th International Conference on Network and System Security.
AI SECURITY · REINFORCEMENT LEARNING
Md. Rakibul
Hassan.
Exploring intelligence.
Building trustworthy security.
Computer Science graduate from BRAC University, focused on AI-driven cybersecurity, reinforcement learning, and adaptive security systems.
Dhaka, Bangladesh / BRAC University
RESEARCH FOCUS
Learning to adapt.
Designed to defend.
INTELLIGENCE × SECURITYCONCEPTUAL NETWORK
3.94/4.00
B.Sc. Computer Science · BRAC University
2
First-authored conference papers
9
Latest research news Consecutive merit scholarships
LATEST / NEWS
Recent research
milestones.
Two first-authored conference papers advancing adaptive cybersecurity and safer Bangla language models.
01 / ABOUT
At the intersection of
learning & security.
I’m a Computer Science graduate with a focus on making security systems more intelligent, adaptive, and robust.
My work explores reinforcement learning for automated network penetration testing, safety in Bangla language models, and the human side of privacy and security.
I developed NASimEmu+, an IDS-aware dynamic network penetration-testing framework using deep hierarchical reinforcement learning for policy learning in defended, noisy, and large-scale networks.
02 / RESEARCH & PUBLICATIONS
Questions into
contributions.
From adaptive network defense to safer language models.
DYNAS
A Dynamic Network Penetration Testing Framework Using Deep Hierarchical Reinforcement Learning
An IDS-aware framework that uses hierarchical reinforcement learning (Pen-DHRL) to learn penetration-testing policies in dynamic, defended network environments.
20th International Conference on Network and System Security
Hangzhou, China · 30 October–1 November 2026
Read preprint Proceedings forthcoming
REPORTED RESULTS
~64%
Higher reward per step
Bridge scenario
~30%
More captured hosts
VarB scenario
Compared with the next-best baseline. Dynamic: ~24% higher reward per step; VarA: ~26% more captured hosts.
BanglaVeilGuard
Cross-Script Safety Benchmarking and Lightweight Guardrails for Bangla Large Language Models
A Bangla-focused safety benchmark spanning Bangla, Romanized Bangla, and code-mixed prompts, with lightweight guardrails to detect and reduce unsafe outputs in low-resource settings.
4th International Conference on Computing Advancements
AIUB, Dhaka · 15–16 October 2026
Read preprint Proceedings forthcoming
UNDERGRADUATE THESIS · DEFENDEDBuilding the foundation: NASimEmu-enhanced
Automated Network Penetration Testing Framework Using Reinforcement Learning
- Led the extension of NASimEmu for realistic, configurable penetration-testing environments.
- Implemented dynamic v2 scenario generation with per-episode stochastic sampling, variable subnet sizes, and randomized host, service, and process configurations.
- Introduced host-state variability and controlled stochastic noise to reduce overfitting and improve policy robustness.
- Refined episode regeneration and the scenario-loading lifecycle while maintaining stable step-level behavior.
- Built reproducible workflows for training, tracing, debugging, and evaluating agents, alongside tooling to compare static and dynamic conditions.
03 / SELECTED PROJECTS
Ideas, explored.
Systems, built.
Experiments, research proposals, and software.
01EXPERIMENTAL STUDY
Experimental Analysis
of DAgger
Comparative experiments and ablation studies of DAgger, Behavior Cloning, and SMILe on handwritten-word recognition and CartPole-v1 control.
Reached a 92% CartPole success rate versus 47% for Behavior Cloning, while OCR accuracy improved from 76.16% to 79.45%.
02RESEARCH PROPOSAL
Privacy, identity
& belonging
A proposed qualitative HCI study of privacy risks and information-sharing practices among Urdu-speaking Bihari communities in Bangladesh.
Examines linguistic stigma, camp life, contested citizenship, and historical marginalization in social media privacy and digital identity management.
03RESEARCH PROPOSAL
One Code,
Many Voices
How pretext shapes OTP disclosure in Bangladesh. A survey and interview study investigating how scam scenarios influence victim compliance.
Compares event-based and fabricated pretexts to understand why users comply with or resist OTP disclosure attempts.
04CRYPTOGRAPHY WORKBENCH
CipherScope
A learning workbench for classical, symmetric, and public-key cryptography, including AES, DES, RSA, and ECC.
Brute-force and factorization demonstrations, frequency analysis, and algorithm performance comparisons.
05FULL-STACK APPLICATION
Share-Notes
An academic note-sharing platform connecting users, courses, and learning materials.
Includes note management, search, payments, and messaging modules.
06WEB APPLICATION
PhD Application System
An academic application management platform built with Laravel.
Database-backed application records and web-based filtering workflows.
07FULL-STACK APPLICATION
Hire-Wizard
A job portal with dedicated recruiter and job-seeker roles.
Job postings, applications, profiles, skill records, and JWT-based authentication.
08MACHINE LEARNING
Prediction of
Obesity Levels
A classification model that predicts obesity categories from eating habits, physical conditions, and lifestyle features.
09COMPUTER GRAPHICS
Space-Station-Siege
A 3D computer graphics game with player movement, shooting mechanics, and real-time game logic.
04 / EDUCATION & EXPERIENCE
A foundation
for what’s next.
Learning deeply.
Sharing what I learn.
B.Sc. in Computer Science
BRAC University · Dhaka, Bangladesh
3.94 / 4.00 CGPA
Algorithms · Computer Networks · Cryptography · Artificial Intelligence · Operating Systems
Merit Scholarship
9 consecutive semesters
Dean’s & VC’s List
Qualified with GPA ≥ 3.7
Undergraduate Teaching Assistant
Discrete Mathematics · BRAC University
Mentored 80+ students through problem-solving sessions and office hours. Prepared quiz solutions and graded quizzes and exams covering 25% of the course grade.
Supervisor: Mr. Sakib Anwar Rieyan
05 / TECHNICAL TOOLKIT
Tools behind
the work.
From learning algorithms
to the systems around them.
- 01 Programming
- Python · JavaScript · R
- 02 Machine learning
- Scikit-learn · PyTorch (basic) · Reinforcement learning concepts
- 03 Networks & security
- Wireshark · Kali · NS3 · Packet Tracer
- 04 Web & databases
- Flask · Laravel · React · MySQL · SQLite
- 05 Foundations & graphics
- Algorithms · Data structures · Cryptography · Assembly (EMU8086) · OpenGL
06 / GET IN TOUCH
Let’s start
a conversation.
AI security, reinforcement learning,
and the questions in between.
ELSEWHERE & CONTACT
GitHub LinkedIn Google Scholar Download curriculum vitae +880 1320499508Dhaka, Bangladesh
Academic references
Dr. Swakkhar Shatabda
Professor, Department of Computer Science and Engineering
BRAC University
Dr. Muhammad Iqbal Hossain
Associate Professor, Department of Computer Science and Engineering
BRAC University