Biometric Authentication Using Ovonic Threshold Switches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric authentication systems require extensive training data and significant computational resources, and they often involve insecure data transmission over communication networks.
Innovation Solution
A biometric authentication and identification system utilizing a quantization unit, volatile threshold switches, an address decoder, a pulse generator, a readout unit, and a database, which processes user-specific biometric time series by assigning quantization indices and selecting threshold switches based on these indices to generate a query vector for authentication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If large but accurate neural networks are trained and deployed on servers with especially equipped hardware, then identification accuracy is improved, but device complexity and energy consumption increase
Solution Approach 1:
The patent replaces complex neural network computations with a resistive memory-based system that uses physical resistance states to store and process biometric data. The Ovonic Threshold Switches (OTS) utilize phase transitions in chalcogenide materials to create high and low resistance states, eliminating the need for traditional CPU/GPU-based neural network processing while maintaining identification accuracy.
Solution Approach 2:
The system changes the operational parameters from digital computation to analog resistance-based storage. By mapping quantized biometric data points to resistance states in memory cells, the system transforms the identification problem into a physical state comparison task, reducing computational complexity while preserving accuracy.
2Measurement precision
If extensive amounts of training data are used, then model accuracy is improved, but training time and computational resources increase
Solution Approach 1:
The patent performs quantization of biometric data during the data collection phase, transforming continuous biometric signals into discrete data points that can be directly mapped to resistance states. This preliminary processing eliminates the need for extensive iterative training, as the system can directly compare quantized stored data with quantized input data through resistance state matching.
3Adaptability or versatility
If biometric data is transmitted over communication networks, then remote authentication is enabled, but security vulnerabilities increase
Solution Approach 1:
The patent creates a local copy of the biometric data processing capability by storing quantized biometric data points in resistive memory on the authentication device itself. This eliminates the need to transmit raw biometric data over networks, as only the processed resistance state representations are stored locally and used for comparison, reducing security risks associated with data transmission.
4Device complexity
If data points of time sequences are mapped to only a fraction of available compute elements, then device complexity is reduced, but computational efficiency decreases
Solution Approach 1:
The patent designs the resistive memory array such that each memory cell can serve multiple functions: storing individual biometric data points, representing quantized values, and participating in parallel comparison operations. The crossbar array architecture allows all memory cells to be simultaneously accessed and utilized during authentication, maximizing computational efficiency without increasing device complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system enables efficient and secure user identification and authentication with minimal training data, low power consumption, and reduced vulnerability to identity theft and data leaks, as it processes biometric data locally on edge devices.
Implementation Method 1
Each threshold switch of the plurality of threshold switches transitions from a high resistance state to a low resistance state when subjected to a voltage pulse exceeding a threshold voltage
Implementation Method 2
Cycling history-dependent shifts in the threshold voltage of each threshold switch are relaxed in the high resistance state
Implementation Method 3
The readout unit is configured to measure a threshold voltage-dependent leakage current of each threshold switch in the high resistance state
Data Source
Figure 1
Figure 2~5
Figure 6~7
AI summary
A biometric authentication and identification system (100) and related methods are disclosed. The system (100) comprises a quantization unit (102) for assigning a quantization index k to each datapoint of a biometric time series, a plurality of volatile threshold switches (108a), an address decoder (105) for selecting, at each time step t of the time series, a single threshold switch of through an index combination comprising at least the quantization index k[t] corresponding to the time step t and the quantization index k[t-T] corresponding to an earlier time step t-T, a pulse generator (115) for applying write pulses to the selected threshold switches, a readout unit (109) for accumulating measured leakage currents of each threshold switch in the high resistance state as components of a query vector, and a database (111) for matching the query vector with at last one biometry class vectors stored as token for a predefined user group. Each threshold switch transitions from a high resistance state to a low resistance state while subjected to a write pulse, and cycling history-dependent shifts in the threshold voltage being relaxed in the high resistance state.