Machine Learning & Mission-Critical Systems
U.S. Ph.D. (Embry-Riddle, 2025) in machine learning security. Five first-author papers, three in IEEE Access. Twelve years on mission-critical systems.
Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Field Notes
Field notes on ML security, proof-of-learning, distributed systems, and high-reliability software for researchers, engineers, and technical leaders.
Posts by Category
Posts organised by category: machine learning, security, engineering, and more.
Posts by Collection
Browse all publications, talks, and portfolio items on ozgurural.github.io.
Essays
Slow, indirect writing about machine learning, security, and the systems we are quietly building.
Blockchain-Enhanced Machine Learning, animated
What a ledger buys machine learning: consensus that trains instead of hashing, contracts that pay for the improvement you caused, and the measured limits.
Detecting Cyber-Security Events in a Low-Resource Language, animated
Detecting cyber-security events with no labelled corpus: a keyword vector learned from the nic.tr attack, Turkish morphology, and anomalies per entity.
Determinism at 60 Hz, animated
What hard real time costs in a Level D simulator: the 150 ms gate, why mean frame time is the wrong statistic, and how latency composes across a rack.
Gradient Pinball: how machines learn, animated
A cinematic, PhD-level explainer on gradient descent, momentum, and why high-dimensional loss landscapes are ruled by saddle points, not local minima.
Universal Jira: Prediction Markets for Global Coordination, animated
A cinematic explainer on how smart contracts and prediction markets can act as incentive-compatible coordination with no trusted central party.
ML Oracles: Verifiable Claims for the Chain, animated
Can a contract trust an AI answer it did not compute? An animated comparison of zero-knowledge proofs, optimistic challenges, and their limits.
Proof-of-Learning (SecurePoL), animated
How Proof-of-Learning checks training records, where spoofing attacks exploit it, and how SecurePoL combines trajectory and watermark verification.
Block Race: Nakamoto consensus, animated
Why ‘6 confirmations’ is a probability, not a promise. A cinematic, PhD-level walk through the double-spend math of Bitcoin’s whitepaper §11.
Redundancy Reactor: fault tolerance, animated
Majority voting buys superlinear safety, until correlation installs a floor you cannot vote past. An animated explainer ending in the Ariane 5 loss.
The Watermarking Wars: Capacity vs. Robustness, animated
A cinematic explainer comparing ML watermarking strategies: parameter perturbations, feature triggers, generative green-lists, and auxiliary heads.
Model Heist Detector: AI watermarks, animated
How a watermark too faint to see in any single weight can become detectable across thousands of them under an explicit statistical model. A cinematic Z-test explainer.
Research Lab
Block Race, Model Heist Detector, SecurePoL, Redundancy Reactor, Gradient Pinball, ML Oracles and more. Real research math, animated from the equation up.
Perspectives & Leadership
Essays on system architecture, engineering leadership, and what twelve years of shipping mission-critical software teaches about resilient systems.
Projects
AI and ML projects by Dr. Ozgur Ural spanning model provenance, verifiable inference, autonomous agents, safety-critical systems, and human-verified clinical tools.
Publications
Peer-reviewed publications by Dr. Ozgur Ural on proof-of-learning, model watermarking, adversarial robustness, and blockchain-enhanced machine learning.
Advisory & Service
Technical advisory, ML security and model provenance review, conference speaking, peer review, and program-committee service.
Sitemap
A full list of all posts, pages, and publications on ozgurural.github.io.
Privacy and Site Use
How this portfolio handles optional analytics, browser preferences, external services, and contact messages.
Posts
Generative Audio Watermarking: Why AI Music Production Needs SecurePoL-Style Provenance Before It Scales
What 4 years of Proof-of-Learning and model-watermarking research tell us about the missing regulatory and security layer in generative AI audio: verifiable provenance for every second of synthesized output, not just for model weights.
Flight Simulators Are Becoming the AI Proving Ground
What if flight simulators are the AI proving ground nobody is talking about? Three years inside Avion.
How to Recognize Advice That Actually Helps
Entrepreneurs, researchers, and engineers live in a torrent of guidance. Podcasts, newsletters, and mentors offer conflicting prescriptions, each delivered w…
The Real Test for Making Something People Want
Scaling conversations dominate startup culture, yet the first question any product must answer is painfully small: will even one person use it when given the…
Why Great Engineering Teams Run on Curiosity
Working on simulator software, defensive cybersecurity platforms, and doctoral research has taught me that durable progress rarely starts with a perfect road…
The Myth of the Quick Fix
Every engineer knows the temptation of the “five-minute hack.” A bug appears, the schedule is tight, and a clever shortcut promises to save the day. Weeks la…
Documentation as the First User Interface
Before a developer ever sees your landing page, they might encounter your README. For many technical products, documentation is the first user interface. It …
Hiring Generalists: The Small Team’s Secret Weapon
Why early-stage engineering teams should hire for breadth and learning rate, and how generalists become the connective tissue that later scales into specialists.
The Invisible Cost of Shiny Tools
Engineering teams love new tools. The promise of faster development, cleaner abstractions, or a more elegant stack is hard to resist. Yet every adoption deci…
Cockpit Discipline for Engineering Teams: What Competitive Sailing on the Aegean Taught Me About Leading Mission-Critical Systems
Six years racing with METU Sailing Club across Urla, Bodrum, and Marmaris taught me seven engineering-leadership patterns I now carry into every program: from Havelsan DLP and Comodo secure gateways through Avion Level-D simulators and Embry-Riddle SecurePoL research.
Why Early Beta Feedback Matters More Than Launch Day
Product launches often steal the spotlight, but the feedback gathered during early beta testing quietly determines whether launch day is a victory lap or a s…
Fixed-Point Invariants in Biology and Machine Learning: Why the Y Combinator Shows Up in Both DNA Replication and Secure Checkpoint Hashing
Why the Y combinator’s fixed-point recursion pattern is the shared computational invariant behind three systems I have built for three separate clients: SecurePoL checkpoint hashes, a 2023 clinical-AI EKG annotation pipeline, and a 2018 Havelsan DLP’s iterative classification loop.
Proof of Learning: Building Trust in Future Machine Learning
Proof of Learning (PoL) verifies that a model was genuinely trained on claimed data by providing verifiable evidence of the training process. I first felt th…
Building Secure AI Supply Chains: From Flight Simulator Architecture to Proof-of-Learning Protocols
How 11 years shipping mission-critical systems (Havelsan DLP, Comodo secure gateways, Avion Level-D simulators) frame the design of spoofing-resilient proof-of-learning protocols at Embry-Riddle’s Cybersecurity & Assured Systems lab.
Model Watermarking and the Future of Trustworthy AI
Model watermarking embeds identifiable patterns into a model’s parameters or outputs so that ownership can be demonstrated without access to the original tra…
ML-Enhanced Blockchain: Toward Intelligent, Adaptive Ledgers
Machine learning (ML) is increasingly used to make blockchain networks more secure, efficient, and user-friendly. When I co-authored our survey on blockchain…
JIRA Meets Smart Contracts: A Production Oracle Architecture for Cross-Border Engineering Governance
The on-chain/off-chain oracle architecture I built for the 2025 Avion-ERAU cross-border SecurePoL delivery program, replacing manual status emails with signed JIRA-to-Smart-Contract oracles that govern milestone payments automatically.
Blockchain-Enhanced Machine Learning: Linking Trust, Data, and Incentives
How blockchain’s immutable ledgers and decentralized governance address auditability, provenance, and incentive-alignment challenges in machine-learning pipelines.
Autonomous Flight Systems: Engineering Mission-Critical Autonomy from Tactical UAVs to Verifiable AI
Architectural lessons from engineering mission-control and ground-control software for Kargu and Togan UAVs: hard real-time constraints, deterministic safety envelopes, and the boundary between learned models and physical actuators.
Engineering Leadership Across Two Regulatory Regimes: Lessons from U.S. Aerospace and Dutch Industrial Systems
How six years split between a Dutch flight-simulator house (Avion, Leiden) and a U.S. aerospace PhD program (Embry-Riddle, Daytona) shapes my playbook for leading regulated, cross-border engineering teams that deliver under both FAA and EASA frames.
Adversarial Example Attacks Against Proof-of-Learning: A Reproduced Threat Reference for Every SecurePoL Defense
The runnable reference implementation of four spoofing attacks against plain Proof-of-Learning verification, used as the calibration threat model for every SecurePoL defense published in IEEE Access 2024, 2025, and the 2025 Embry-Riddle doctoral dissertation.
SecurePoL with Watermarking: The Jupyter Notebook Suite Behind Three IEEE Access Papers and a European FFS Deployment
The official repository of Jupyter notebooks implementing the three watermarking strategies compared across the SecurePoL line of work, with the measured spoofing-resistance numbers reported in the IEEE Access 2024 and 2025 papers and the 2025 Embry-Riddle doctoral dissertation.
Clinician-Calibrated EKG Measurement with PyQt5: The Deterministic Baseline for a Clinical-AI Annotation Pipeline
A PyQt5 desktop tool for clinician-validated EKG interval measurement, used to generate 312 signed, verified labels that now form the ground-truth baseline for a 2026 Turkish-hospital AI-assisted landmark-detection pilot. No neural components; purely deterministic measurement logic for regulatory traceability.
Presenting at ICISSP 2021: Why Low-Resource, Agglutinative Languages Break Standard Security Detectors, and What We Measured
The in-person talk at the International Conference on Information Systems Security and Privacy 2021, co-authored with Prof. Cengiz Acartürk, covering the agglutinative-morphology failure mode and the operational false-positive budget that made the Turkish cyber-event detector usable.
Turkish Cyber-Security Event Detection: The Master’s Thesis That Became a Production Threat-Intelligence Feed
The 2019 METU M.S. thesis building an agglutinative-language NLP pipeline for detecting cyber-security events from Turkish Twitter streams and newspaper archives: the measured results, the publication at ICISSP 2021, and how the morphological-normalization module now ships inside a European CTI feed product.
Secure Proxy on Cloud: A 2016 Architecture for Cloud-Native Security Policy Enforcement, and What Became of It
The 2016 ICESC conference paper and demonstration video documenting a cloud-native secure proxy architecture, the design patterns that survived five years into the Comodo Secure Web Gateway product line, and the lessons that now shape Avion Level-D simulator policy planes.
publications
Autonomous Cargo and Mail Delivery Robot (Clover)
Published in Turkish Autonomous Robots Conference (Otonom Robotlar Konferansı, Ankara), 2014
2014 METU undergraduate capstone project and conference paper on an indoor autonomous cargo and mail delivery robot. The perception-inside-deterministic-safety-envelope architectural pattern demonstrated here is the same structure applied five years later to the STM Kargu and Togan autonomous UAV programs....
Recommended citation: Ural, O., and the Clover Capstone Team (2014). Autonomous Cargo and Mail Delivery. Proceedings of the Turkish Autonomous Robots Conference, Ankara, Turkey.
Read paper
Secure Proxy on Cloud
Published in International Conference on Engineering and Security (ICESC 2016), 2016
2016 ICESC conference paper and demonstration documenting a cloud-native four-module secure proxy architecture. The policy-enforcement separation, streaming-normalization, and signed-audit patterns first published here now ship in the 2018 Comodo Secure Web Gateway product line and in Avion Level-D simulator data-egress...
Recommended citation: Ural, O. (2016). Secure Proxy on Cloud. Proceedings of the International Conference on Engineering and Security (ICESC 2016). DOI: 10.13140/RG.2.2.24058.08649.
Read paper
Automatic Detection of Cyber Security Events from Turkish Twitter Stream and Turkish Newspaper Data
Published in Master's Thesis, Middle East Technical University (Ankara, Turkey), 2019
2019 METU M.S. thesis introducing a morphological-normalization and per-entity anomaly-scoring pipeline for Turkish cyber-security event detection. The morphological-normalization module is now licensed and deployed in a European MSSP’s Turkish-language OSINT feed, serving 80+ managed-security customers.
Recommended citation: Ural, O. (2019). Automatic Detection of Cyber Security Events from Turkish Twitter Stream and Turkish Newspaper Data. Master's Thesis in Cyber Security, Middle East Technical University. Advisor: Prof. Cengiz Acartürk.
Read paper
Automatic Detection of Cyber Security Events from Turkish Twitter Stream and Newspaper Data
Published in Proceedings of the 7th International Conference on Information Systems Security and Privacy (ICISSP), 2021
NLP-based pipeline for automated detection of cybersecurity incidents from Turkish Twitter and news streams using TF-IDF and ensemble classifiers.
Recommended citation: Ural, O. and Acartürk, C. (2021). "Automatic Detection of Cyber Security Events from Turkish Twitter Stream and Newspaper Data." In Proceedings of the 7th International Conference on Information Systems Security and Privacy (ICISSP), pp. 66-76. DOI: 10.5220/0010201600660076.
Read paper
Survey on Blockchain-Enhanced Machine Learning
Published in IEEE Access, 2023
Comprehensive survey of blockchain-enhanced machine learning: consensus-driven data provenance, federated learning on-chain, and incentive mechanisms across 120+ papers.
Recommended citation: Ural, O. and Yoshigoe, K. (2023). Survey on Blockchain-Enhanced Machine Learning. IEEE Access, pp. 145331-145362. DOI: 10.1109/ACCESS.2023.3344669.
Read paper
Enhancing Security of Proof-of-Learning against Spoofing Attacks using Feature-Based Model Watermarking
Published in IEEE Access, 2024
Feature-based model watermarking scheme that binds ownership proofs to internal activations, surviving fine-tuning and transfer attacks on Proof-of-Learning.
Recommended citation: Ural, O. and Yoshigoe, K. (2024). Enhancing Security of Proof-of-Learning against Spoofing Attacks using Feature-Based Model Watermarking. IEEE Access. DOI: 10.1109/ACCESS.2024.3489776.
Read paper
Enhancing Proof-of-Learning Security Against Spoofing Attacks Using Model Watermarking
Published in Doctoral Dissertation, 2025
Doctoral research developing SecurePoL, a dual-layer framework coupling Proof-of-Learning trajectory logs with three orthogonal watermarking strategies.
Recommended citation: Ural, O. (2025). Enhancing Proof-of-Learning Security Against Spoofing Attacks Using Model Watermarking. Doctoral dissertation, Embry-Riddle Aeronautical University.
Read paper
SecurePoL: Integration of Watermarking With Proof-of-Learning to Enhance Security Against Spoofing Attacks
Published in IEEE Access, 2025
Dual-layer framework coupling immutable Proof-of-Learning logs with three watermarking strategies, so verification succeeds only when both the training trajectory and the watermark are consistent.
Recommended citation: Ural, O. and Yoshigoe, K. (2025). SecurePoL: Integration of Watermarking With Proof-of-Learning to Enhance Security Against Spoofing Attacks. IEEE Access, vol. 13, pp. 213067-213091. DOI: 10.1109/ACCESS.2025.3642198.
Read paper