Recycled IC Detection via PSRR Degradation and ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting recycled counterfeit electronics, such as integrated circuit chips (ICs) and system-on-chips (SoCs), are inadequate due to the lack of a universal, inexpensive, and automated technique that does not require golden samples or extensive hardware modifications.
Innovation Solution
A method utilizing machine learning models based on the degradation of power supply rejection ratio (PSRR) data from low drop-out regulators (LDOs) to classify ICs and SoCs as recycled or new, which involves processing PSRR data using a recycle detection machine learning model to determine if a suspect component is recycled.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If targeted electrical testing with golden samples is used to detect recycled ICs, then detection accuracy is improved, but the requirement for golden samples and increased device complexity worsens
Solution Approach 1:
The patent extracts the detection function from complex hardware modifications and golden sample requirements, isolating the essential detection mechanism to PSRR measurement of the LDO power supply. This extraction eliminates the need for golden samples while maintaining detection capability through the inherent aging characteristics of the power supply circuitry.
Solution Approach 2:
The patent creates a universal detection method that works across different IC vendors and types by focusing on the common LDO power supply component. The PSRR measurement approach is vendor-agnostic and can detect recycled components from any manufacturer without requiring vendor-specific golden samples or customized test equipment.
2Reliability
If physical inspection with X-ray tomography is used to detect counterfeits, then detection capability is improved, but cost and device complexity worsen
Solution Approach 1:
The patent replaces complex mechanical imaging systems (X-ray tomography, electron microscopy) with an electrical measurement system that assesses PSRR characteristics. This substitution transitions from expensive, complex physical inspection infrastructure to simpler electrical testing that leverages the natural electrical aging signatures of recycled components.
3Reliability
If hardware security primitives like PUFs are designed into new chip architectures, then cloned counterfeit detection is improved, but device complexity and manufacturing cost worsen
Solution Approach 1:
The patent copies the detection approach from specialized security primitives to the existing LDO power supply circuitry. Instead of adding dedicated security hardware, it replicates the detection functionality using the already-present PSRR measurement capability of the LDO, which naturally exhibits aging characteristics that can be measured and compared.
4Productivity
If automated testing procedures are implemented across all electronics types, then productivity is improved, but adaptability worsens due to the wide variety of counterfeit components
Solution Approach 1:
The patent achieves both automation and universality by focusing on the LDO power supply component that is present in virtually all electronic devices. The automated PSRR measurement procedure can be applied uniformly across different device types, vendors, and technologies because the LDO power supply is a universal building block in electronics, providing both automation efficiency and broad adaptability.
Data Source
AI summary
Embodiments of the present disclosure provide methods, systems, apparatus, and computer program products are for detecting whether a suspect component such as an integrated circuit (IC) or a system-on-chip (SoC) is recycled. Specifically, various embodiments involve processing power supply rejection ratio (PSRR) data obtained from a low drop-out regulator (LDO) used for the suspect component using a recycle detection machine learning model to generate a recycle prediction. In particular embodiments, the recycle detection machine learning model is developed based at least in part on degradation of PSRRs of LDOs. Accordingly, a determination is made as to whether the suspect component is recycled based on the recycle prediction. If so, then an indication that the suspect component is recycled is provided.