Josemaria
Osuorah
Cybersecurity researcher · Systems engineer · Author and mentor
Cybersecurity researcher · Systems engineer · Author and mentor
I am a PhD researcher at Prairie View A&M University, working where machine learning meets cybersecurity. The question I keep coming back to is how systems stay secure at scale without giving up the performance they were built for.
My work covers security for networks and IoT deployments. Protocol-layer weaknesses in low-power wireless systems, adaptive machine learning adversaries against hardware-rooted authentication, provenance across edge-to-cloud data paths. Most of it gets built and tested on real hardware sitting on a bench rather than only in simulation, which is slower but tends to surface the problems that actually matter.
Before this I spent years in industry, supporting enterprise Azure and Microsoft 365 environments at Tek Experts and working on telecommunications infrastructure at Nigeocom for Huawei projects across major mobile networks. That shaped how I think about research. Security work that ignores operational reality tends to stay in the paper, and engineering that ignores the research tends to repeat mistakes someone already documented.
I came to Houston from Nigeria, and a good part of my time outside the lab goes into helping people making a similar move find their footing in security and engineering.
JADA uses a reinforcement learning agent, trained with Proximal Policy Optimization, to decide which Physical Unclonable Function challenges carry the most learning value under a restricted query budget. Random Forest models then learn device behaviour from what it collects.
The finding matters for how PUF security gets evaluated. Measured against a passive attacker, PUF strength looks better than it is. An attacker that chooses strategically changes the threat model.
LoRa and similar LPWAN technologies are built for long range and minimal power draw, and that lightweight design leaves exposure across the whole protocol stack. My survey work maps those weaknesses layer by layer rather than treating the stack as one surface.
A provenance-aware LoRaWAN platform built on physical hardware, combining Heltec end devices, a Dragino gateway, ChirpStack, MQTT, containerised services and Python processing, to verify device identity, packet lineage and payload integrity from sensor to cloud.
An automated, fault-tolerant data acquisition framework for the Prairie View A&M Microgrid Testbed, supporting continuous edge-to-cloud telemetry and secure storage of energy-system measurements for monitoring and analytics.
A practical introduction to digital security covering network protection, encryption, malware, access control and safe online practice, with equal weight given to the human behaviour that most breaches actually turn on.
On contemporary network security and evolving cyber threats, with practical approaches to protecting modern digital systems. Written for practitioners working with systems already in production.
I started the Secure Mentorship Network to help people find a way into cybersecurity and solutions engineering, especially those without an obvious route in. Most of it is unglamorous. Reading CVs, explaining what a job posting is actually asking for, talking someone through their first time on call.
I also review papers for IEEE Access and for the SATC and ISBCom conferences, which is its own kind of mentorship and a good way to stay honest about my own writing.
Notes on cloud security, identity and the operational side of the research. Mostly things I worked out the hard way and wrote down so the next person does not have to.
Read the blogAlways glad to hear about research collaborations, speaking, or anything security related worth talking through. Email reaches me fastest.