TinyML Security and Privacy Training in Madrid, Spain

The TinyML Security & Privacy Training focuses on securing machine learning models and data on ultra-constrained edge devices such as microcontrollers and embedded IoT systems.
Participants learn how to protect models, firmware, data, and device integrity, while ensuring privacy, compliance, and trust in TinyML deployments used in industrial, consumer, healthcare, and critical infrastructure environments.

This course bridges embedded security with ML-specific threats, helping teams build secure-by-design TinyML systems.

Introduction to TinyML Security & Privacy

• Why security and privacy matter in TinyML

• Differences between cloud AI and TinyML security

• Unique risks of resource-constrained devices

• Security-by-design for embedded AI

Threat Landscape for TinyML

• Physical device attacks

• Firmware and model extraction

• Adversarial inputs on edge devices

• Data poisoning and manipulation risks

• Side-channel and fault-injection concepts

Device Identity & Trust

• Unique device identity

• Secure provisioning and onboarding

• Root of trust concepts

• Certificate and key management basics

Secure Boot & Firmware Integrity

• Secure boot principles

• Firmware signing and verification

• Preventing unauthorized firmware

• Rollback protection and version control

Protecting TinyML Models

• Model encryption and obfuscation

• Preventing model theft and IP leakage

• Secure storage of model weights

• Tamper detection strategies

Data Privacy on TinyML Devices

• Minimizing sensitive data collection

• On-device processing vs cloud upload

• Data anonymization and filtering

• Privacy-by-design for edge AI

Secure Inference & Runtime Protection

• Protecting inference pipelines

• Memory protection and isolation

• Preventing runtime manipulation

• Detecting abnormal behavior

Adversarial ML at the Edge

• Adversarial examples (conceptual)

• Physical-world adversarial attacks

• Robustness techniques for TinyML

• Monitoring for abnormal inputs

Secure Communication & Edge-to-Cloud Privacy

• Encrypting data in transit

• Secure APIs and messaging

• Authentication between edge and cloud

• Preventing data leakage

Secure Updates & Model Lifecycle

• Secure over-the-air (OTA) updates

• Model versioning and rollback

• Patch management for TinyML

• Preventing downgrade and replay attacks

Compliance, Governance & Ethics

• Data protection regulations (conceptual overview)

• Privacy impact assessments

• Audit trails for embedded AI

• Responsible AI in edge deployments

Incident Response & Recovery

• Detecting compromised devices

• Containment and remediation strategies

• Forensics basics for embedded systems

• Recovery and re-provisioning

Hands-on Exercises

Summary and Conclusion