This course was developed as a MEST Micro-Certificate. MEST Micro-Certificates offer short, self-paced learning opportunities through focused training modules and webinars on specialized technical topics. Participants who successfully complete the program requirements earn a certificate of completion.
These programs are designed for U.S. citizens, including Department of War and other government personnel, government-affiliated employees, industry professionals, college students, and faculty.
The LLM Based SoC Security Verification II micro-certificate introduces students to advanced AI- and LLM-assisted techniques that build upon the foundations established in LLM Based SoC Security Verification I. This micro-certificate focuses on automating key hardware security verification tasks, including threat modeling and security test plan generation, LLM-driven synthesis of security-focused RTL testbenches, and systematic detection of RTL-level security bugs. Through hands-on, agentic workflows, learners gain practical experience in scalable, evidence-driven verification of modern SoC designs.
This course covers three major LLM-driven steps for hardware security verification:
Unit 1: Threat Modeling and test Plan Generation using LLM: This lab training module teaches how to use Large Language Models in an agentic, retrieval-augmented framework to automate hardware security threat modeling and generate structured, design-aware security test plans. Participants learn to analyze physical, supply-chain, and software-exploitable threats by extracting security policies from hardware specifications and ISA documents, and translating them into actionable verification tasks using multi-agent reasoning and human-in-the-loop refinement.
Unit 2: LLM based Testbench Generation for Bug Detection: This lab teaches how to use a multi-agent, LLM-driven framework to automatically generate security-focused RTL testbenches that detect and validate hardware vulnerabilities. Students learn how LLMs analyze RTL designs and CWE descriptions to form bug hypotheses, generate temporally precise test scenarios, and synthesize executable SystemVerilog testbenches that trigger real security flaws in simulation.
Unit 3: Hardware Bug Detection with LLM: This lab module teaches how Large Language Models can be applied to automatically detect security-critical bugs in RTL finite state machine designs, reducing reliance on manual and rule-based verification. Participants learn to use the SecRT-LLM framework to analyze Verilog code and identify structural vulnerabilities such as Hamming distance violations, deadlocks, unreachable states, and duplicate state encodings through guided prompting and fidelity checking.
Unit 4: Supplimentry Lecture: This unit features an expert-led presentations on how emerging AI capabilities, particularly LLMs operating as intelligent agents, can transform hardware security verification.
Format: Online, asynchronous
Platform: nanoHUB Learning Management System
Estimated Time: 18-24 hours
Level: Advanced Undergraduate/Graduate Student
Prerequisites:
Basic understanding of RTL design, SoC architecture, and hardware security concepts
Familiarity with SystemVerilog testbench, threat modeling, and hardware level bugs
Prior exposure to LLM frameworks (RAG, and prompting) or EDA verification flows is desired but not required.
MEST Micro-Certificate: Introduction to Large Language Models (LLMs) (recommended)
MEST Micro-certificate: System-on-Chip (SoC) Verification (recommended)
Instructors: Professor Mark M. Tehranipoor, University of Florida & Professor Farah Farahmandi, University of Florida
Course Objectives:
Explain how agentic LLM architectures overcome the limitations of manual and rule-based hardware security verification, particularly for threat modeling, RTL vulnerability detection, and test generation.
Use retrieval-augmented and in-context LLM pipelines to extract security policies, attack knowledge, and CWE-aligned vulnerability descriptions from hardware specifications, ISAs, and RTL designs.
Identify and classify hardware security threats across physical, supply-chain, architectural, and RTL-level attack surfaces using LLM-driven reasoning and user-guided context refinement.
Apply LLM-based analysis to detect structural RTL security bugs, including FSM deadlocks, unreachable states, Hamming distance violations, and duplicate state encodings.
Generate structured security verification artifacts using LLMs, including threat models, bug hypotheses, security policies, test objectives, and temporally aligned test scenarios.
Synthesize and execute security-focused verification collateral, such as SystemVerilog testbenches and simulation-based tests, to validate LLM-identified vulnerabilities with runtime evidence.
Evaluate the trustworthiness and limitations of LLM-generated security insights by correlating model explanations with simulation outcomes, fidelity checks, and known real-world hardware failure cases.
Request Course Enrollment
To enroll in this course, you must have a nanoHUB account. Create one here: https://nanohub.org/register/
Request membership in the MEST group in nanoHUB at https://nanohub.org/groups/mest