Counterfeit semiconductors have always been a problem during good times and bad.
During the most recent bad time in the supply chain — when COVID-19 lockdowns caused a massive chip shortage — counterfeit chips soared as companies took desperate measures to source the electronic components they needed for designs.
To combat counterfeit semiconductors, researchers at Purdue University’s College of Engineering have developed an optical detection method called residual attention-based processing of tampered optical responses (RAPTOR) that uses deep learning to identify tampering.
Purdue said the platform improves on current detection methods that face challenges in scalability and discriminating between natural degradation and adversarial tampering.
“Our scheme opens a large opportunity for the adoption of deep learning-based anti-counterfeit methods in the semiconductor industry,” said Alexander Kildishev, professor at Purdue University.
While several techniques have been developed to affirm semiconductor authenticity and detect counterfeit chips, these have largely used physical security tags baked into the chip functionality or packaging.
“However, there are significant challenges in achieving scalability and maintaining accurate discrimination between adversarial tampering and natural degradation, such as physical aging at higher temperatures, packaging abrasions and humidity impact,” Kildishev said.
How RAPTOR works
“RAPTOR is a novel deep-learning approach, a discriminator that identifies tampering by analyzing gold nanoparticle patterns embedded on chips,” Kildishev said. “It is robust under adversarial tampering features such as malicious package abrasions, compromised thermal treatment and adversarial tearing.”
The system uses the distance matrix verification of gold nanoparticles that are randomly and uniformly distributed on the chip sample substrate. Gold nanoparticles can easily be measured using dark-field microscopy and is a technique that can be integrated into any stage of the semiconductor fabrication pipeline.
“RAPTOR uses an attention mechanism for prioritizing nanoparticle correlations across pre-tamper and post-tamper samples before passing them into a residual attention-based deep convolutional classifier,” said Blake Wilson, an alumni at Purdue University working on the project. “It takes nanoparticles in descending order of radii to construct the distance matrices and radii from the pre-tamper and post-tamper samples.”
Testing the system
The Purdue team tested the capabilities of RAPTOR by simulating the tampering behavior in nanoparticle systems that includes:
- Natural changes
- Malicious adversarial tampering
- Thermal fluctuations
- Varying degrees of random Gaussian translations of the nanoparticles
“We have proved that RAPTOR has the highest average accuracy, correctly detecting tampering in 97.6% of distance matrices under worst-case scenario tampering assumptions,” Wilson said. “This exceeds the performance of the previous methods — Hausdorff, Procrustes and Average Hausdorff distance — by 40.6%, 37.3%, and 6.4%, respectively.”
The next steps are to collaborate with chip-packaging researchers to further innovate the nanoparticle embedding process and streamline authentication steps, Purdue said.
The full research can be found in the journal SPIE Digital Library.
