Detect IoT attacks in progress, with fewer false alarms.
100% on-device detection. Your traffic never leaves your network.
Results from doctoral research on a labeled public test dataset. See details and limitations.
Traditional security tools can't keep pace with IoT expansion
Connected IoT devices are projected to grow from about 10 billion in 2019 to about 30 billion by 2030.1
Attackers increasingly target edge and network devices that “typically lack standard endpoint detection and response (EDR) telemetry.”2
Hybrid AI-Powered IoT Threat Detection
LSTM (supervised) + Isolation Forest (unsupervised)
Evaluated on normal traffic plus DDoS, DoS TCP flood, port scanning, ICMP flood and other anomaly classes
In doctoral research, the hybrid model cut the false-positive rate from 0.38% (standalone LSTM) to 0.30%, about 20% fewer false alarms
Minimal footprint: runs on a single edge device, no cloud dependency.
ZNET AI Defender builds on Dr. Akere's doctoral research. Development continues independently at A4C Labs, the R&D lab of A4C Prod LLC.
The hybrid LSTM + Isolation Forest model was built and evaluated using Microsoft Azure cloud compute. It was trained on a labeled public IoT traffic dataset (Stratosphere Laboratory, via Kaggle) and tested on a held-out subset. The test covered normal traffic, DDoS, DoS TCP flood, port scanning, ICMP flood, and other anomalies.
The model now runs on an NVIDIA Jetson Orin Nano at A4C Labs and analyzes live traffic from a growing set of real consumer and smart-home IoT devices. This phase tests what the research couldn't: performance on unlabeled, real-world traffic.
Performance figures describe research results on a specific labeled test dataset and are not a guarantee of results in any particular environment.
Python, Keras (LSTM) and scikit-learn (Isolation Forest); edge inference on NVIDIA Jetson Orin Nano
Planned support for Pcap, MQTT, Modbus, and custom IoT protocols
Planned SHAP-based feature attribution so analysts can see why traffic was flagged
Manufacturers that supply the defense supply chain run connected sensors, PLCs, and controllers next to systems that handle Controlled Unclassified Information. Most of those devices can’t run an endpoint agent, and CMMC Level 2 still expects you to monitor them. ZNET AI Defender watches their network behavior instead.
Healthcare IoT, energy and utilities, and smart infrastructure: any environment with devices that can’t run security agents.
Built for environments where devices can’t run agents and traffic shouldn’t leave the site
| Buyer concern | Gap we address | ZNET AI Defender approach |
|---|---|---|
| Hardware cost | Network detection often means dedicated appliances or sensors | Edge inference on a ~$250 NVIDIA Jetson Orin Nano |
| Model transparency | Hard-to-inspect “black box” scoring | Inspectable LSTM + Isolation Forest models with published research results |
| Data locality | Traffic or telemetry sent off-site for analysis | On-site detection; in the current pilot, traffic stays local |
| Cost to adopt | Enterprise pricing out of reach for smaller manufacturers | Low-cost edge hardware, sized for small and mid-size sites |
Expertise in cybersecurity, AI/ML, and research
Founder, Developer, CISO & CEO
Cybersecurity engineer and vCISO with 20+ years in risk and compliance. U.S. Navy veteran. University instructor in intrusion detection and forensics, cryptography, and network security. CISSP, CISA, PCI-QSA, CDPSE, HITRUST CCSFP, CySA+. D.Eng in Cybersecurity Analytics (GWU).
Request a demo or contact us to learn how ZNET AI Defender detects anomalous IoT and OT traffic
Seeking $250K in pre-seed funding to ship a commercial MVP and run 3 design-partner pilots.
The investment is in ZNET AI Defender, a product of A4C Prod LLC, which holds its intellectual property.
Investment Inquiry: invest@znet.ai