AI-powered optical detection could tackle fake chips

Researchers in the US have proposed a method of protecting semiconductor chips from counterfeiting that relies on deep learning techniques and an optical marker.

The team from Purdue University say the technology – called RAPTOR (Residual, Attention-based Processing of Tampered Optical Responses) – identifies tampering by analysing gold nanoparticle patterns embedded on chips.

They describe the approach in the journal Advanced Photonics, saying it can protect against key methods of chip counterfeiting based on 'adversarial tampering' – in other words an attempt to monitor or affect the correct operation of a chip or a security core within it.

That includes "malicious package abrasions, compromised thermal treatment, and adversarial tearing," according to the researchers, who say their technique can distinguish between tampering and natural degradation of a chip caused by ageing or exposure to temperature or humidity changes.

"The semiconductor industry has grown into a $500bn global market [but] is grappling with dual challenges: a profound shortage of new chips and a surge of counterfeit chips, introducing substantial risks of malfunction and unwanted surveillance," according to a statement on the research published by Purdue.

"Those factors have contributed to a $75bn counterfeit chip market that "jeopardises safety and security across multiple sectors dependent on semiconductor technologies, such as aviation, communications, quantum, artificial intelligence, and personal finance."

Current methods deployed to protect chips from counterfeiting typically rely on physical security tags baked into the chip functionality or packaging, but the Purdue team's approach relies on physically unclonable functions (PUFs) – a term for markers based on random processes that are deemed almost impossible to replicate.

RAPTOR was trained on a 10,000-image dataset of randomly distributed gold nanoparticles and was shown to be 97.6 per cent effective at detecting tampering, a more robust result than other PUF approaches, according to the authors of the paper.

"The ease of fabrication of gold nanoparticles, along with rapid and robust tampering detection with RAPTOR, opens up a large opportunity for the adoption of machine-learning-based tampering detection schemes in the semiconductor industry," they conclude.