Artificial intelligence is now taking center stage in the fight against one of the most invisible yet dangerous threats to modern technology — hardware trojans. A research team at the University of Missouri has unveiled an AI-powered detection system that can identify hidden malicious code in computer chips with up to 97% accuracy, offering a groundbreaking solution for safeguarding global semiconductor supply chains.
From smartphones and medical devices to defense systems, microchips are the unseen backbone of our digital world. Yet, deep inside these chips, hardware trojans—tiny, malicious alterations in chip design—pose serious risks. They can steal sensitive data, weaken system security, and even sabotage national infrastructure. Detecting these trojans has traditionally been a slow, complex, and costly process, but researchers at Missouri have developed a smarter, faster, and more transparent approach.
Led by doctoral candidate Ripan Kumar Kundu from Mizzou’s College of Engineering, the project leverages large language models (LLMs) — the same AI technology that powers advanced chatbots — to scan chip design files for anomalies. The method not only flags suspicious patterns in the design but also provides clear explanations of why the code might be malicious. “That explanation is critical because it saves developers from digging through thousands of lines of code,” said Kundu. “We’re making the process faster, clearer, and more trustworthy.”
Their system, called PEARL (Predictive Explainable AI for Reliable Logic), is detailed in the paper “An Adaptive and Explainable Hardware Trojan Detection Using Open Source and Enterprise Large Language Models,” published in IEEE Access. What makes PEARL unique is its adaptability — it can run on local systems or cloud platforms, making it practical for both open-source developers and enterprise manufacturers. This flexibility means the technology could easily integrate into global chip production across industries like healthcare, finance, consumer electronics, and defense.
Why it matters: Unlike software malware, hardware trojans cannot be deleted or patched once the chip has been manufactured. They can remain dormant until triggered, at which point they may cause data leaks, system failures, or even critical national security threats. Given the global and multi-layered nature of the chip production process, vulnerabilities can be introduced at nearly any stage. The Mizzou team’s approach helps detect and eliminate threats early, reducing risks of costly product recalls or supply chain disruptions.
Khurram Khalil, co-author and fellow doctoral researcher, emphasized the broader implications: “These chips are the foundation of our digital world. By combining AI with explainability, we’re ensuring that foundation remains secure at every level.” The research group is also exploring real-time automated repair mechanisms, which could enable chips to self-correct vulnerabilities before deployment.
In conclusion, the University of Missouri’s innovation marks a major leap in AI-driven cybersecurity for hardware. By fusing advanced machine learning with explainable analysis, the PEARL system promises to make the world’s semiconductor ecosystem more resilient, transparent, and secure — ensuring the technology we rely on daily remains trustworthy from design to deployment.
