CYBERSECURITY RESEARCHER

Hamizan Azman

Security research at Singapore Management University, supervised by Prof Xiaofei Xie. First-author paper on LLM agent frameworks in the works. CVE-2026-27855 published in Dovecot at CVSS 6.8, plus three further bounties on HackerOne at CVSS 6.9, 6.8, and 4.8.

Singapore Management University
NTUCyberSG R&D Programme Office
Temasek PolytechnicDigital & Intelligence Service
Google.orgThe Asia Foundation

01 // ABOUT

I'm a 19-year-old cybersecurity researcher in Singapore, recently graduating with a Diploma in Cybersecurity and Digital Forensics at Temasek Polytechnic.

Since February 2026 I have been a research intern at Singapore Management University under Prof Xiaofei Xie, with Dr Lili Quan as day-to-day supervisor. First author on a paper proposing a bug-class taxonomy and fix-pattern design for LLM agent frameworks, targeting a top security or software-engineering venue.

I founded the TP Cybersecurity Clinic as its first Lead Student Ambassador from August 2025 to February 2026, running Singapore's first polytechnic-led cybersecurity clinic for MSMEs. I oversaw 38 MSMEs being supported and 40 ambassadors trained, with 12 on-site engagements I personally led. I was behind its admittance into the CyberSG Consortium.

On the disclosure side, I have one CVE published at CVSS 6.8 Medium, a second bug at CVSS 6.9 Medium, third at CVSS 6.8 Medium, and a fourth at CVSS 4.8 Medium. Four paid bug bounties total across three programmes, six valid-but-duplicate findings, and currently active in an invite-only GovTech Bug Bounty Programme.

I was accepted into the DIS University Work-Learn Scheme as a Cyber Specialist for 2026.

LocationSingapore
EducationTP Cybersecurity & Digital Forensics
GraduatedMay 2026
RoleResearch Intern @ SMU
SupervisorProf Xiaofei Xie
FocusLLM Security
DIS UWLSCyber Specialist (2026)
CertificationCompTIA Security+

02 // RESEARCH

FIRST-AUTHOR PAPER

Bug-Class Taxonomy for LLM Agent Frameworks

I'm first author on a paper mapping the kinds of bugs that show up in real LLM agent frameworks (LangChain, LangGraph, CrewAI, LlamaIndex) and proposing a small library of fix patterns developers can reach for. We treat reproducibility under model non-determinism as its own axis, which prior work (Zhang, Zhu, "When Agents Fail") does not. Supervised by Prof Xiaofei Xie and Dr Lili Quan.

4

Frameworks Studied

1st

Author

LLM AgentsLangChainLangGraphCrewAILlamaIndexSoftware Engineering
LIB2APP CONTRIBUTOR

Open-Source LLM Application Deployments

Work on Lib2App, led by Yuelin Wang (Tianjin University PhD), on reproducibility of open-source LLM applications. Triaged 102 apps and got 79 deploying as pinned Docker images, with 83 published on Docker Hub under hoomzoom. The 23 that wouldn't deploy each have a written failure analysis (mostly Python or CUDA version mismatches and upstream submodule rot).

102

Apps Triaged

79

Deployed

83

Docker Images

DockerPythonDependency PinningDocker HubWSL2
RESEARCH

TrustChain PenTest Engine

I help build the lab's automated application pentest engine led by a PhD researcher at SMU. It runs as 26 Docker services behind one entry port, doing exploit generation, CVE search, DOCX report writing, and SSE progress streaming. The platform exposes per-tool progress at stage 1 and a partial-automation mode with rerun. Source is private.

26

Microservices

8

Tasks Delivered

DockerMicroservicesLLM AgentsCVE SearchReport Generation
LIB2APP CONTRIBUTOR

Vulnerability PoC Reproduction Framework

Lib2App work continued. Reproduced 231 vulnerability PoCs from the lab's library-CVE dataset as Dockerised shooting ranges. Each PoC bundles a vulnerable server, attack script, Dockerfile, README, and reproduction notes. Anyone can clone and run.

231

PoCs Reproduced

100%

Completion Rate

DockerPythonFlaskHTTP ExploitationVulnerability Research
FINAL-YEAR PROJECT

Forward Proxy Security

Final-year project at Temasek Polytechnic on forward proxy security. A demanding architecture-level problem that took me deep into network-layer threat modelling and TLS interception.

Y3

FYP

Network SecurityForward ProxyTLSArchitecture

03 // DISCLOSURES

Active vulnerability research in open-source authentication and protocol code. One CVE published, three further bounties on HackerOne with CVEs pending assignment, and currently active in an invite-only GovTech programme.

1

CVE Published

Dovecot CVE-2026-27855 (CVSS 6.8 Medium)

3

CVEs In Process

CVSS 6.9, 6.8, and 4.8 via HackerOne, bounties paid

4

Paid Bounties

Across 3 programmes

6

Valid-but-Dupes

Sustained research practice

Published

CVE-2026-27855 (Dovecot)

CVE-2026-27855 (Dovecot)

auth_cache_remove() uses the wrong username field, allowing OTP replay when passdb rewrites the username during improper authentication. CVSS 6.8 Medium. Monetary reward via YesWeHack. Acknowledged on Dovecot's security disclosures page.

Rewarded, CVE pending

Three further HackerOne findings

Three further HackerOne findings

Three further bug bounties rewarded via HackerOne, at CVSS 6.9 Medium, CVSS 6.8 Medium, and CVSS 4.8 Medium. CVE identifiers pending vendor coordination.

Active

GovTech Bug Bounty Programme 17 (GBBP17)

Invited into and currently active in GBBP17, the GovTech Bug Bounty Programme. Invite-only. Programme details under NDA.

Ongoing

Coordinated disclosure pipeline

Further open-source disclosures in progress alongside a collaborator. Details withheld until reports are public.

04 // EXPERIENCE

LLM Security Research Intern

Singapore Management University

Selected via NTU CRPO Cyber Translation Internship Programme

FEB 2026 - PRESENT

Research intern under Prof Xiaofei Xie, with Dr Lili Quan as day-to-day supervisor. First author on a paper proposing a bug-class taxonomy and fix-pattern design for LLM agent frameworks across LangChain, LangGraph, CrewAI, and LlamaIndex.

  • First-author paper proposing a bug-class taxonomy and fix-pattern design for LLM agent frameworks (LangChain, LangGraph, CrewAI, LlamaIndex). Supervised by Prof Xiaofei Xie and Dr Lili Quan.
  • Lib2App contributions (paper led by Yuelin Wang): triaged 102 open-source LLM and AI applications. 79 are reproducible Docker images with pinned dependencies. 83 are public on Docker Hub under hoomzoom.
  • Lib2App contributions: reproduced 231 vulnerability PoCs as Dockerised shooting ranges. Anyone can clone and run.
  • Helping build the lab's automated pentest engine of 26 Docker services. Covers exploit generation, CVE search, DOCX report generation, and SSE progress streaming.
AI/LLM SecurityPrompt InjectionMCPSupply Chain SecurityDockerVulnerability Research

Lead Student Ambassador

TP Cybersecurity Clinic

Sponsored by The Asia Foundation, supported by Google.org

AUG 2025 - FEB 2026

Founding Lead Student Ambassador with pre-pilot involvement before the August 2025 launch. Ran Singapore's first polytechnic-led cybersecurity clinic for micro, small, and medium enterprises.

  • Conducted 12 on-site cybersecurity engagements personally, the most of any ambassador
  • Recruited, trained, and mentored 40 ambassadors
  • Clinic supported 38 MSMEs across Singapore, with three companies returning for up to four engagements
  • Two MSMEs offered me internships on the spot during engagements
  • Represented the Clinic in monthly online meetings with peer cyber clinics at UT Austin, UC Berkeley, and other global institutions
  • Drove the Clinic's admittance into the CyberSG Consortium and ongoing collaboration with NTU CRPO
LeadershipCybersecurity ConsultingCommunity ImpactTraining

Malware Analyst Intern

TP Malware Analysis Centre

MAY 2025 - AUG 2025

Researched deepfake detection methods. Represented Temasek Polytechnic at national and overseas events. Also conducted independent malware reverse engineering.

  • Selected to present DeepVysion+ (TP 2024-cohort Best Major Project) at GovWare 2025 to an international audience
  • Selected to present DeepVysion+ at SWITCH 2025 (Marina Bay Sands). A booth interview with an NTU CRPO researcher led directly to the SMU research internship offer.
  • Selected to present DeepVysion+ to David Neo, Major-General (MG) Lee Yi-Jin, and other SAF and DIS staff at SAF Day 2025
  • Presented at TP MAC to Singapore Police Force, Lifelong Learning Institute, NUS School of Computing leadership, and multiple overseas delegations
Malware AnalysisReverse EngineeringDeepfake DetectionPublic Speaking

05 // EVENTS

GovWare 2025

Presenter, DeepVysion+

Selected to present DeepVysion+, the Temasek Polytechnic 2024-cohort Best Major Project, to an international cybersecurity audience.

SAF Day 2025

Presenter, DeepVysion+

SAF Day 2025

Selected to present DeepVysion+ to David Neo, Major-General (MG) Lee Yi-Jin, and other SAF and DIS staff.

TP Malware Analysis Centre visits

Presenter

Selected to present DeepVysion+ and lab capabilities at TP MAC to visiting groups including Singapore Police Force, Lifelong Learning Institute, and NUS School of Computing leadership.

Common ICT Taster Programme 2025

Student speaker and coordinator

Spoke to the entire Temasek Polytechnic freshman cohort in a theatre setting. Coordinated and led other senior students who presented across cybersecurity and IT modules.

TP Open House

Speaker

Spoke as a representative during Temasek Polytechnic's Open House.

TP x HK IIT Inaugural International Hackathon 2025

Volunteer staff

Volunteer staff supporting the inaugural Temasek Polytechnic x Hong Kong IIT International Hackathon.

06 // RECOGNITION

  • 2026

    DIS University Work-Learn Scheme

    Digital and Intelligence Service (Singapore)

    Accepted as a Cyber Specialist for the 2026 intake, after passing IPPT Silver in December 2025.

  • 2025

    PolyFinTech100 API Hackathon (NETS track)

    Singapore FinTech Festival

    Runner-up. One spot away from the prize.

  • 2025

    AMP Education Bursary

    AMP Singapore

    Bursary recipient. Same body funded my CompTIA Security+ certification voucher.

  • 2024

    GovTech AI Hackathon

    GovTech Singapore

    Top 20% of participants.

  • 2024

    SparkCTF

    Singapore Polytechnic

    Participated.

07 // COMMUNITY

  • Jun 2024 - Present

    Malay Activity Executive Committee, Punggol Community Club

    Volunteer on the MAEC. Helped run multiple Hari Raya events, food and care distributions, and community get-togethers.

  • Sep 2025 - Present

    Homage care work

    Daily care shifts on Homage across Singapore. Mixed schedule of short visits, overnight, and full-day shifts. Have not missed a day except for declared holidays and illness.

  • Mar 2026

    Tarawih4Youth at Masjid Sultan

    Staff member during the Tarawih4Youth programme at Masjid Sultan, Ramadan 2026.

  • Feb 2026

    Ramadan 2026 food distribution

    Helped distribute food to the needy during Ramadan 2026.

  • Nov 2025

    Secondary 4 mentoring

    Hosted a Secondary 4 student for a one-week shadow programme at Temasek Polytechnic. Taught throughout the week and motivated him to begin HTB CPTS and Cisco CCNA. He is now around halfway through both at age 15, and sent a thank-you email to TP lecturers about the mentorship.

08 // CERTIFICATIONS

CompTIA Security+ SY0-701

CompTIA

Completed

Industry standard cybersecurity certification. Funded by AMP Singapore.

Certified 06 July 2025Expires 06 July 2028Code ZM78MC1S0EB15REJVerify
CompTIA Security+ SY0-701

HTB Certified Penetration Testing Specialist (CPTS)

Hack The Box

In progress

Sponsored by AMP Singapore.

10 // TESTIMONIAL

“I recommend him without reservation and am confident he will continue to excel in any academic or professional setting he chooses to pursue. He is a capable leader, a reliable team member, and a young professional of integrity and promise.”

Mehreen Tanvir

Program Manager, The Asia Foundation

Recommendation Letter Page 1Recommendation Letter Page 2

11 // IMPACT

1

CVE Published

Dovecot CVE-2026-27855, CVSS 6.8 Medium

3

CVEs In Process

CVSS 6.9, 6.8, and 4.8 via HackerOne, bounties paid

4

Paid Bug Bounties

Across 3 programmes

6

Valid-but-Dupes

Sustained research practice

231

Vulnerability PoCs Reproduced

Full dataset rebuilt as runnable shooting ranges

79

LLM Apps Containerised

83 pinned images on Docker Hub under hoomzoom

38

MSMEs Supported

Through the TP Cybersecurity Clinic

40

Ambassadors Trained

Recruited and mentored as Lead Student Ambassador

12

On-Site Engagements

Personally led, the most of any ambassador

12 // CONTACT

Let's connect.

I'm open to research collaborations, mentorship opportunities, and conversations about AI/LLM security. Currently based in Singapore.

19 · TP Cybersecurity & Digital Forensics graduate · focus on LLM security

Built by Hamizan Azman · hamizanazman.com