I'm Mustafa Akpinar, an Information Systems & AI graduate student working at the intersection of security engineering, threat intelligence, and LLM research. I turn complex security problems into measurable, practical work.
01 / SELECTED WORK
Original research, real-world security experience, and hands-on engineering. The most relevant projects come first.
Designed a Python benchmark comparing three LLMs and three prompting strategies on the BigVul vulnerability dataset.
View case study ↗A separate MITRE internship project focused on cyber deception and defensive detection concepts. Specific implementation details are not published.
View case study ↗Implemented Microsoft Defender deception capabilities and helped connect deception alerts with XDR-based investigation workflows.
View case study ↗Explore additional security and software projects.
Built Splunk dashboard visualizations to centralize Microsoft Defender deception alerts for SOC monitoring.
Explore case study ↗Researched Keycloak-based identity federation and IAM risks spanning Entra ID, cloud, and on-premises environments.
Explore case study ↗Applied authenticated firmware, rollback protection, ARM memory isolation, and access-control testing in a technical security challenge.
Explore case study ↗Built a Java Android app using an MNIST TensorFlow Lite model with PIN setup, verification, and lockout logic.
Explore case study ↗Designed a five-department, 200-user network concept with VLANs, firewall policies, VPN access, and managed wireless connectivity.
Explore case study ↗Internship projects are deliberately presented as high-level public-safe case studies. No proprietary code, internal architecture, or operational secrets are published.
02 / RESEARCH NOTE
My independent study investigated whether language models can consistently identify vulnerable code. The benchmark makes failure modes visible—not just headline accuracy.
Read the research case study ↗Zero-shot, few-shot, and CWE-reference context tested with reproducible sample sets.
Prompting strategies had model-specific effects; added reference material did not consistently help.
Correctly naming a weakness type proved harder than binary vulnerability identification.
03 / EXPERIENCE
Practical security experience connecting investigation, detection, and engineering to business decisions.
Evaluated cyber threat intelligence capabilities, studied federation and IAM risks, and developed scripting and networking diagnostics in lab contexts.
Investigated threats, supported vulnerability prioritization, and contributed to enterprise endpoint deception and security-monitoring initiatives.
04 / ABOUT
My interests span the questions that connect cybersecurity research to real operational outcomes: how to evaluate tools, secure modern infrastructure, understand threats, and communicate technical risk clearly.
Outside technical work, I care about mentorship, collaboration, and connecting people across different backgrounds. I see communication as part of good security work—not an afterthought.
M.S. Information Systems & Artificial Intelligence · Expected 2028
B.S. Information Systems · Cybersecurity concentration · 2026
05 / TOOLKIT
SOC analysis, threat intelligence, vulnerability assessment, security investigation, detection workflows, MITRE ATT&CK, NIST
Python, PowerShell, Java, SQL, model benchmarking, prompt evaluation, structured experiments, quantitative analysis
Splunk, Microsoft Defender XDR, Recorded Future, Google Threat Intelligence, Keycloak, Entra ID, Azure, AWS, VMware, TCP/IP
06 / CONNECT
Open to conversations about security consulting, applied security research, threat intelligence, and engineering.