Medical terminal trusted authentication method based on physically unclonable function

Abstract

The application discloses a medical terminal trusted authentication method based on a physically unclonable function and relates to the technical field of terminal trusted authentication. The method comprises the following steps: a medical terminal receives a reference challenge vector and synchronously collects a physiological electrical signal and a body temperature plaintext; a dynamic internal mask is extracted from the least significant bit of a continuous sampling window to generate thermal shot noise and a degradation flag is output; a cross switch network is driven by the dynamic internal mask to perform same-weight bit shuffling, a transient challenge vector is formed, an SRAM-PUF is injected to obtain a noisy initial response, and the noisy initial response, the body temperature plaintext and the degradation flag are reported; then, a verification server performs thermal compensation on a digital twin baseline according to the body temperature plaintext to generate a predicted response and a contraction coefficient; finally, the tolerance threshold is tightened according to the contraction coefficient, blind search comparison is completed in a limited reverse de-shuffling space, and when the degradation flag is valid, the method is switched to a static mask isolation area to perform physical fuse verification.

Description

Technical Field

[0001] This invention relates to the field of terminal trusted authentication technology, specifically a trusted authentication method for medical terminals based on physically unclonable functions. Background Technology

[0002] For wearable or implantable medical devices, bedside terminals with sensing capabilities, and diagnostic and treatment equipment requiring continuous network connectivity, the terminals need to collect physiological parameters, monitor local conditions, and transmit data, while also connecting to hospital platforms, cloud servers, or supporting control equipment. Current practices involve generating response or key materials by calling physically unclonable functions on the terminal side, and utilizing processors, registers, communication interfaces, and error correction or auxiliary data mechanisms to complete local storage, remote comparison, or key derivation. When temperature is involved, temperature sampling, temperature gauge entries, and authentication also need to be invoked simultaneously to reduce the impact of environmental changes.

[0003] US Patent document US20210194707A1 discloses a temperature-sensing physically non-clonable function authentication system. It uses conductive components such as resistors, capacitors, and on-chip wires as the physical model for the security token. A two-bit key is output through paired resistors and an analog comparator. An analog-to-digital converter obtains the electrical output of one of the nodes. A processor connects to a key register and a temperature meter to store the key bits, temperature, and timestamp, and outputs authentication or temperature measurement information to an external system via a communication interface. This scheme encapsulates related components within a single package to reduce component exposure and replacement. From an operational perspective, its principle is to transform the electrical differences caused by the discreteness of device manufacturing into a physically non-clonable function output, using electrical components with a temperature coefficient to generate temperature readings, thereby achieving security tokens and temperature sensing within the same hardware system.

[0004] The aforementioned technologies can integrate temperature sampling and physically unclonable function (PHF) authentication on the same hardware, but their core remains the construction of a temperature-sensing security token and maintaining the token's output availability. For scenarios like medical terminals that operate continuously, sample continuously, and experience significant dynamic changes in their working environment, the terminals are affected not only by temperature but also by factors such as sampling link disturbances, device aging, power supply fluctuations, and external interference. Existing physically unclonable functions generally require the introduction of error correction, auxiliary data, or reconstruction mechanisms.

[0005] Existing literature also indicates that some applications of physically non-cloning functions require error correction to reconstruct stable outputs, and that SRAM physically non-cloning functions and other implementations can be modeled, cloned, or improperly implemented, increasing the attack surface. Therefore, in medical terminal scenarios, if the authentication link relies on stable reconstruction, entry compensation, or conventional challenge response, systematic drift and transient noise can easily couple when environmental temperature changes and operational disturbances occur simultaneously. This can lead to inconsistent authentication results for legitimate devices or force the system to relax judgment conditions and increase the link's exposure surface.

[0006] Therefore, the problem that needs to be solved in this field is: how to balance the consistency of the physical non-clonable function authentication output and the security of the authentication link when temperature changes and dynamic disturbances occur simultaneously during the operation of medical terminals. Summary of the Invention

[0007] (a) Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a trusted authentication method for medical terminals based on physically non-clonable functions. This method utilizes a dynamic internal mask to drive a cross-connected network to perform equal-weighted shuffling, forming a transient challenge vector which is then injected into a SRAM-PUF to obtain a noisy initial response. Subsequently, a verification server performs thermal compensation on the digital twin baseline based on plaintext body temperature, generating a predicted response and a shrinkage coefficient. Finally, based on the shrinkage coefficient, a tolerance threshold is tightened, and blind search comparison is completed within a limited reverse deshuffling space. When the degradation flag is valid, the system switches to a static mask isolation zone to perform physical fuse verification, thereby reducing temperature drift interference and the risk of auxiliary data exposure.

[0009] (II) Technical Solution

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A trusted authentication method for medical terminals based on physically unclonable functions includes: after receiving a baseline challenge vector, the medical terminal simultaneously collects physiological electrical signals and body temperature plaintext; extracts thermal shot noise from the least significant bit of a continuous sampling window to generate a dynamic internal mask and outputs a degradation flag; the medical terminal uses the dynamic internal mask to drive a cross-switching network to perform bit-weighted shuffling on the baseline challenge vector, generates a transient challenge vector and injects it into an SRAM-PUF; after obtaining a noisy initial response, the medical terminal reports the noisy initial response, body temperature plaintext, and degradation flag.

[0012] The verification server compensates the digital twin baseline based on the plaintext body temperature to obtain the predicted response and shrinkage coefficient. It then tightens the tolerance threshold based on the shrinkage coefficient and performs blind search comparison of the noisy initial response within a limited reverse demixing space. Alternatively, it switches to the static mask isolation area to perform physical fuse verification when the degradation flag is valid.

[0013] Furthermore, after receiving the benchmark challenge vector, the medical terminal drives the physiological electrical signal sampling channel and the body temperature sampling channel with the same hardware clock, obtains the corresponding sampling sequence and body temperature plaintext within the same authentication cycle, and extracts the least significant bit thermal shot noise within the continuous sampling window to generate a dynamic internal mask and simultaneously determine the degradation flag.

[0014] Furthermore, the medical terminal determines the starting offset of the continuous sampling window according to the benchmark challenge vector, extracts binary results from the target bit plane, and generates a dynamic internal mask according to the predetermined extraction step size and folding order. The total length of the dynamic internal mask is consistent with the total number of control bits of the cross switch network. After generation, it resides in the local hardware only during this round of certification.

[0015] Furthermore, the medical terminal writes the dynamic internal mask into the inter-stage control register of the cross-connect network, and each level of the switching unit performs in-situ pass-through or position swapping on adjacent challenge bits in the reference challenge vector, outputting a transient challenge vector. The same-weight bit shuffling only changes the bit position arrangement of the reference challenge vector, without changing the Hamming weight sum of the reference challenge vector.

[0016] Furthermore, the medical terminal directly sends the transient challenge vector into the word line and bit line addressing chain of the SRAM-PUF. After the target cell is selected, it performs controlled power-on and latches the internal node voltage difference of each memory cell at a fixed sampling time to form a noisy initial response. The authentication message reported by the medical terminal includes the noisy initial response, body temperature plaintext, and degradation flag, but does not include auxiliary data.

[0017] Furthermore, during the registration phase, the verification server establishes a digital twin parameter library for each medical terminal, storing reference temperature, reference mobility, thermal index, bit-level barrier difference, thermal coupling gain, transition width, bit flip threshold, and reference sign bit. During the authentication phase, the plaintext body temperature is substituted into the thermodynamic equation to obtain the bit flip trend under the current thermal state, and the digital twin baseline is thermally shifted accordingly to generate a predicted response.

[0018] Furthermore, the verification server calculates a shrinkage coefficient that characterizes the degree of residual misalignment after temperature drift stripping, based on the bit-level misalignment relationship between the noisy initial response and the predicted response, and in combination with the bit weights determined for each response bit during the registration phase. The verification server associates and stores the shrinkage coefficient with the uncompensated misalignment result under the reference state as a direct input for subsequent tightening of the tolerance threshold.

[0019] Furthermore, the verification server reads the reference tolerance threshold saved during the registration phase and generates the tightening tolerance threshold for the current session based on the shrinkage coefficient. Subsequently, within the finite reverse deshuffling space corresponding to the same weight shuffling, the noisy initial response is compared concurrently, the misalignment of each candidate branch relative to the predicted response is calculated, and the existence of a unique hit branch is determined by the degree of separation between the smallest misaligned branch and the second smallest misaligned branch.

[0020] Furthermore, when the downgrade flag is valid, the verification server stops the regular authentication path and sends an isolation routing command to the medical terminal; the medical terminal closes the main array communication path and switches to the static mask isolation zone, reads the isolation pattern corresponding to the isolation routing command, forms an emergency verification message, and returns it to the verification server.

[0021] Furthermore, the static mask isolation zone contains multiple pre-registered isolation patterns, and each isolation pattern is associated with a corresponding physical fuse unit. When the verification server confirms that the emergency verification message is consistent with the isolation registration database, the medical terminal applies a fuse pulse to the physical fuse unit called for this authentication, causing the corresponding isolation pattern to become invalid in subsequent authentication sessions.

[0022] (III) Beneficial Effects

[0023] This invention provides a trusted authentication method for medical terminals based on physically non-clonable functions, which has the following beneficial effects:

[0024] Within the same authentication cycle, the medical terminal generates a dynamic internal mask, plaintext body temperature, and downgrade flag to ensure that the authentication input corresponds to the current physical state of the terminal, thereby reducing the impact of inconsistencies between the authentication input and the actual operating state of the terminal.

[0025] The medical terminal uses a dynamic internal mask to drive the cross-connect network to perform same-weight shuffling and sends the transient challenge vector into a physical non-cloning function array to obtain a noisy initial response. The authentication message does not contain auxiliary data, thereby reducing the risk of auxiliary data exposure in the link.

[0026] The verification server incorporates plaintext body temperature data into the thermal compensation process, thermally shifts the digital twin baseline, and generates a predicted response and contraction coefficient, thereby reducing the impact of systematic thermal drift on authentication decisions.

[0027] The verification server tightens the tolerance threshold based on the shrinkage coefficient and performs concurrent comparisons of noisy initial responses within a limited reverse unmixing space, thereby improving the reliability of the unique hit branch determination; when the degradation flag is valid, it switches to the isolated verification path, thereby improving the authentication credibility in abnormal scenarios. Attached Figure Description

[0028] Figure 1 This is a diagram illustrating the overall architecture of a trusted authentication system for medical terminals.

[0029] Figure 2 The flowchart shows the overall process of a trusted authentication method for medical terminals based on physically unclonable functions.

[0030] Figure 3 To generate a graph for querying traction sampling and dynamic internal masking;

[0031] Figure 4 Generate graphs of raw responses to dynamically internally masked, same-weighted bit shuffling and physically unclonable functions;

[0032] Figure 5 A graph generated by thermal drift stripping and predicted response driven by plaintext body temperature.

[0033] Figure 6 For the shrinking threshold-driven reverse unmixing and concurrent blind search authentication graph;

[0034] Figure 7 The isolation verification path diagram triggered by the downgrade flag. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Please see Figures 1-7 This invention provides a trusted authentication method for medical terminals based on physically non-cloning functions, comprising:

[0037] This method is designed for implantable heart rhythm monitors, insulin pump control terminals, bedside monitoring patches, and portable medical terminals with on-chip memory arrays. When performing trusted authentication on such devices, it is neither suitable to add a separate random number chip nor to send error correction auxiliary data externally for extended periods; otherwise, it would consume power and packaging budget, and expose a learnable physical fingerprint structure.

[0038] More importantly, the terminal acquisition link is constantly exposed to clinical physiological electrical signals such as ECG, pulse wave, and electromyography, and the sampled values ​​are simultaneously affected by body temperature, tissue impedance, motion traction, and external electromagnetic disturbances. The task of step one is not to add another safety chain, but to transform the existing acquisition process into a starting chain of challenge-traction sampling – noise floor folding into a mask – boundary crossing isolation, allowing subsequent steps to obtain the dynamic internal mask in one go. Body temperature readings and downgrade sign .

[0039] Step 1: Obtain the dynamic internal mask for this certification only by analyzing the transient thermal shot noise in the existing sampling front-end of the medical terminal. Output body temperature in plaintext With demotion sign It is used as input for subsequent isotope mixing, thermal drift stripping, and isolation verification.

[0040] When the benchmark challenge vector When the physiological sampling link is mismatched, attackers can form stable observation samples through repeated stimulation; if body temperature is read asynchronously after sampling, the temperature drift compensation on the server side will lose its time correspondence; if pathological extremes or strong electromagnetic injections are not stripped away, the extracted mask will be mixed with violent disturbances unrelated to authentication. Therefore, step one adopts a single-chain approach of challenge-driven sampling, noise extraction, and parallel boundary discrimination. The acquisition controller in the medical terminal uniformly drives the analog front-end, on-chip analog-to-digital converter, and body temperature acquisition branch, enabling dynamic internal masking. Body temperature readings With demotion sign Formed within the same certification cycle.

[0041] The medical terminal receives the baseline challenge vector issued by the verification server. Subsequently, the acquisition controller does not send it to the subsequent storage array, but instead converts it into starting positioning information within a continuous sampling window. The reason for this processing is that the reference challenge vector... Once the sampling method is changed, the digital challenge becomes tied to the analog sampling, and the noise floor extracted subsequently is no longer an isolated random quantity unrelated to the challenge.

[0042] To enable this mapping to be implemented directly, the acquisition controller modifies the baseline challenge vector with a fixed group width. Slice the data and map each slice to the starting offset of a consecutive sampling window:

[0043]

[0044] Among them, the initial offset : No. The starting position of each sampling sub-window within the continuous sampling window, with values ​​ranging from 0 to... Its function is to convert the challenge bits into window positioning results; group width The number of challenge bits used in a single positioning calculation is preferably 4 to 8, which helps to balance positioning resolution and hardware overhead.

[0045] Question Bit : Benchmark challenge vector The Middle The binary value of a bit, taking the value 0 or 1, serves to form the positioning input; the length of the continuous sampling window. The number of consecutive sampling points that can be called in a single authentication is preferably 64 to 512, which serves to limit the boundary of subsequent thermal noise extraction.

[0046] In a representative embodiment, the medical terminal is an implantable heart rate monitor, with platinum-iridium alloy electrode pairs used for physiological data acquisition and an on-chip bandgap temperature sensor used for temperature acquisition. In a parallel embodiment, the medical terminal is a skin-attached monitoring patch, with silver / silver chloride electrodes used for physiological data acquisition and a thermistor and voltage divider circuit used for temperature acquisition. Both types of terminals are driven by the same clock for both the physiological sampling channel and the temperature channel, ensuring that the body temperature is displayed clearly. relative to the starting offset They fall within the same certification cycle.

[0047] When using, the benchmark challenge vector The data was converted into sampling traction data, and subsequent noise floor extraction obtained a clear window source for the body temperature data in plaintext. It also has time consistency that can be used for the next step.

[0048]

[0049] in, Indicates the first Each clinical physiological electrical signal sampling time; This indicates the body temperature sampling time corresponding to this sampling time; Indicates the start time of the current authentication sampling window; This indicates the sampling period of the on-chip analog-to-digital converter; This indicates the maximum permissible time deviation between the physiological sampling channel and the body temperature sampling channel;

[0050] Medical terminals preferably use the same hardware clock source to drive the physiological sampling channel and the body temperature sampling channel, and meet the requirements within the same certification cycle. The body temperature value is recorded as the plaintext body temperature used in this round of certification. .

[0051] After sampling and localization are completed, the on-chip analog-to-digital converter outputs the raw sampled code, which includes thermal noise, quantization boundary jitter, and subtle fluctuations in organizational coupling. Step one does not extract the higher-order amplitude portions; instead, it extracts discrete bits only from the selected least significant bit plane and then folds these discrete bits into a dynamic internal mask. .

[0052] Dynamic internal mask target length The total number of control bits for the cross switch network in step two is preset, where:

[0053]

[0054] in, Indicates the number of switching stages in a crossbar network; Indicates the first Number of level switching units; Its purpose is to ensure the dynamic internal mask output in step one. Each of these corresponds one-to-one with the control register in step two.

[0055] The medical terminal reads the analog-to-digital converter output code in a fixed sampling order within the continuous sampling window. From the pre-selected target location plane Extract bit plane bits with a fixed extraction step size and folding span Generate mask bits Continue accumulating until a length of [length missing] is obtained. Dynamic internal mask .

[0056] If the number of bits generated in the current window is insufficient The terminal continues sampling subsequent windows until the number of bits is filled; if the number of bits generated in the current window exceeds a certain threshold... Then, the first part is truncated according to the generation order. This bit serves as a dynamic internal mask used in this round of certification. Among them, the higher-order bits provide physiological and operational information, while the least significant bit is closer to the current physical fluctuations of the device, and therefore should be a one-time mask.

[0057] Bit-plane extraction uses explicit binarization:

[0058]

[0059] Among them, bit plane bits : No. Each sampling code In the The binary result on the unit plane takes the value 0 or 1. The least significant bit of the analog-to-digital converter output is stripped, and the sampling code is... The analog-to-digital converter outputs the first... A digital quantization value, the range of which is determined by the resolution of the analog-to-digital converter; bit plane number. The bit order from least significant bit to most significant bit is used, with bits 2 to 4 corresponding to the third to fifth least significant bits, balancing transient response and extractability.

[0060] The acquisition controller generates dynamic internal mask bits by fixing the folding span in the security register:

[0061]

[0062] Among them, the mask bits Dynamic internal mask The Middle This bit, with a value of 0 or 1, serves to form the addressing control input for the next step; folding span The number of bit plane bits participating in XOR folding is preferably 3 to 8, which weakens the dominance of single-point mutations. This number is determined by firmware pre-setting. The extraction step size is also specified. The index spacing when selecting bit plane bits is preferably 1 to 4, which is to avoid the mask bits from being completely dependent on adjacent sampling points;

[0063] Position plane number : Actually used to generate dynamic internal masks The target bit plane, the range of values ​​and the bit plane index Consistency, its function is to lock the noise layer used in this round of mask generation; starting offset Taken from the aforementioned starting offset The location result takes values ​​from 0 to... Its function is to make the mask generation path subject to the benchmark challenge vector. Traction; Continuous sampling window length The meaning and range of values ​​are the same as before, and its function is to provide the boundary of the circular index;

[0064] In one specific implementation, the acquisition controller reads the initial offset from the continuous ECG sampling window. For each sampling point, extract the third least significant bit and the fourth least significant bit, and then sort them according to the folding span. Generate dynamic internal mask It is written to an erase-type security register; after the authentication time ends, the register is cleared by hardware without going through the direct memory access channel or being exposed in the debug interface.

[0065] In use, the underlying fluctuations that would otherwise be discarded during medical sampling are transformed into disposable mask material, and the dynamic internal mask... It only stays on the end side for a short time and does not form a long-term exposed sample.

[0066] Dynamic internal mask The sampling link only has authentication value when it remains within a usable physical range. If the terminal experiences severe signal distortion due to malignant arrhythmias, or external electromagnetic injection from electrosurgical units, MRI peripheral equipment, or strong near-field emission sources, the fluctuations in the least significant bit will no longer primarily reflect the device's natural thermal disturbances, but will be mixed with abnormal mutations from pathological or attack events. Therefore, in step one, the boundary deviation index is calculated in parallel during mask generation, and a degradation flag is set based on consecutive boundary violations. .

[0067] Furthermore, the boundary deviation index adopts a combined form of jump density and out-of-band proportion:

[0068]

[0069] Among them, the boundary deviation index : The degree of abnormal deviation within the current sliding window, with a non-negative real number as the criterion for triggering a degradation mechanism; Window length The number of consecutive sampling points used in the discrimination process is preferably between 16 and 128, which serves to limit the time scale for boundary discrimination; sampling code. The meaning and range of values ​​are the same as before; its function is to provide the basis for calculating jump density; jump threshold The difference between adjacent sampled codes is determined as the boundary of abnormal changes. The range of values ​​is set by the front-end quantization bit width. Its function is to isolate normal quantization jitter from abnormal changes.

[0070] Indicator Function : Logical conditional operator, takes a value of 1 when the condition within the parentheses is true, and a value of 0 otherwise; its function is to count abnormal jumps; out-of-band flags The analog front-end monitoring branch outputs a binary value of 0 or 1 to high-frequency interference or saturation events, incorporating electromagnetic injection or front-end distortion information into a single criterion; it also simulates the output of a saturation comparator or high-frequency energy detector at the front end. Weighting coefficients. : Out-of-band labeled terms deviate from the boundary index The contribution ratio in the data is greater than 0 and not higher than 1. Its function is to suppress the excessive dominance of a single source of anomaly in the decision.

[0071] As a supplement: if the boundary deviates from the exponent If the threshold is exceeded in a series of preset sliding windows, the medical terminal will be set to a downgrade flag. Suspend the regular acquisition bus and prevent subsequent samples from being used to generate the dynamic internal mask. .

[0072] When the boundary deviates from the exponent When multiple sliding windows exceed a preset threshold consecutively, the hardware interrupt controller suspends the regular acquisition bus and blocks the dynamic internal mask. Continue refreshing and display the downgrade flag. Set to valid. In a clearly defined on-site implementation, when the implantable heart rhythm monitor performs authentication, a strong near-field electromagnetic source approaches, and the simulated front-end monitoring branch first provides an out-of-band marker. Boundary deviation index After consecutive out-of-bounds errors, the interrupt controller disconnects the shared direct memory access channel, retaining the latched plaintext body temperature data. At the same time, stop writing new samples to the security register, and ultimately only record the body temperature in plaintext. and downgrade sign Sent to the next stage of the process.

[0073] Thus, anomalous sampling is eliminated at its source, and subsequent work will not be performed on a distorted physical base. Benchmark Challenge Vector Once bound as a sampling traction quantity, the medical terminal can simultaneously obtain and query the relevant sampling window and body temperature plaintext within the same authentication cycle. The least significant bit transient thermal shot noise is folded into a dynamic internal mask that is only valid for the current round. Then, the next step is to directly call the internal mask without modifying the random source.

[0074] Step 2: The medical terminal uses the dynamic internal mask output in Step 1. For the benchmark challenge vector Perform shuffling with the same weight to obtain a transient challenge vector that is only valid for this round of authentication. and the transient challenge vector Injecting a physically non-clonable function into the memory array to obtain a noisy initial response. Noisy initial response without generating auxiliary data Body temperature readings and downgrade sign Send to the verification server.

[0075] Dynamic internal mask in step one Using transient thermal shot noise from the medical terminal sampling front end is effective for this round of authentication, but not suitable for storage; benchmark challenge vector The data originates from the verification server, ensuring structural stability and serving as a reproducible external driver across rounds. Simply XORing the challenge structure alters the bit pattern but also changes the Hamming weights, making it impossible for subsequent servers to determine whether errors stem from temperature drift, metastable releases, or changes in the challenge structure. Therefore, this step employs a single-chain operation with equal-weighted bit shuffling, array power-on readout, and direct pass-through of raw responses, along with dynamic internal masking. Only change the position, do not change the total number of bits, body temperature in plaintext. Continue to be tied to the response cycle, downgrade flag If an error occurs, you can immediately switch to the next step.

[0076] The medical terminal receives the dynamic internal mask delivered in step one. Next, first set the dynamic internal mask. The reference challenge vector is latched into the inter-stage control register of the crossbar switch network, and then loaded by the challenge loader. The data are sequentially fed into a multi-level two-input two-output switching unit. Instead of a lookup table-based permutation or software sorting, a hardware-level adjacent switching chain is used. This is because each action of the adjacent switching chain only performs one of two operations: in-situ pass-through or position swap between two input bits. This facilitates close placement on the layout and naturally maintains the Hamming weights.

[0077] For a pair of adjacent query bits, the first Level 1 The output of each switching unit can be expressed as:

[0078]

[0079] Among them, inter-level interrogation positions Inter-level inter-question position They represent the first Entering the first level before the exchange Each exchange unit has two input bits, both taking values ​​of 0 or 1, which are used to form local challenge pairs to be exchanged; inter-level challenge bits Inter-level inter-question position They represent the first The two outputs after the level swap are both 0 or 1, and their function is to be sent to the next swap level or used as a transient challenge vector. Components; control bits : No. Level 1 The switching selection value of each switching unit is 0 or 1. When it is 0, it performs in-situ pass-through; when it is 1, it performs position swapping.

[0080] Level number : Switching level number, a non-negative integer, used to specify the timing level in the switching chain; cell number. : The number of the current level of switching unit, which is a non-negative integer and is used to locate the specific switching position.

[0081] control bit It is not given by a fixed table, but by a dynamic internal mask. Obtained through cyclic extraction, its specific form is as follows:

[0082]

[0083] Among them, the mask bits Dynamic internal mask The value at the specified loop index, which can be 0 or 1, maps the transient secret generated in step one to the specific exchange action; the step coefficient. Dynamic internal masking for different swap levels The index advance amount at that time is preferably a positive integer, which is used to avoid adjacent switching stages from reusing the same mask bit; mask length Dynamic internal mask The total number of bits, which takes the value of a positive integer, is used to provide the boundary for circular addressing.

[0084] In a representative implementation, the crossbar switch network is arranged before the physically unclonable function of the storage array; the switching unit is composed of complementary transmission gates and hold-and-inverter circuits compatible with the terminal master controller; and the control register adopts a load-on-release secure latch structure. (The last part, "when dynamic internal mask," appears to be an error and doesn't need a direct translation.) After writing, the benchmark challenge vector The transient challenge vector is generated from the least significant bit to the most significant bit through a swapping stage. In parallel implementations, the switching unit can also be implemented using a metal programmable interconnect array, as long as each stage output still only performs either in-place or swapped operations, thus maintaining the same weighted bit shuffling principle. In use, a dynamic internal mask is used. For the benchmark challenge vector Its function is limited to changing the position without changing the total amount, and the physical perturbation unique to this round of certification is embedded into the permutation path.

[0085] The cross switch network consists of The system consists of adjacent switching units at level 1, the first... Level includes Each switching unit receives a pair of adjacent input bits and performs one of two actions: in-situ pass-through or position swap. Dynamic internal mask. The The position as the first Level 1 Control bits of each switching unit ,in, For cross-level index step coefficients, This is the length of the dynamic internal mask. Therefore, the total size of the control vector space formed in step two is:

[0086]

[0087] in, This represents the original number of combinations of control vectors. During the deployment phase, the verification server performs offline deduplication on the forward permutations corresponding to all control vectors based on the fixed topology of the crossbar network, obtaining a unique inverse unshuffle set. },in, This represents the total number of unique branches after deduplication. Step four only performs a reverse unmixing and blind search within the set, without expanding the search outside the set.

[0088] After shuffling with equal weights, the medical terminal does not process the transient challenge vector. Instead of being cached by a general-purpose processor, the data is directly fed into the row-column addressing chain of the physically non-cloning function of the memory array. This arrangement is because the transient challenge vector... Its value lies in its interaction with dynamic internal masks. If the data persists concurrently, and remains on the bus or in the cache for too long, it increases the chance of being bypassed and sampled. This step preferably uses a physically non-cloning function for power-on memory arrays: transient challenge vector. First, the selected word line group, bit line group, and readout order are determined. Then, the array enters the controlled power-up phase. The cross-coupled inverters inside the cells collapse from a metastable state to a certain stable state. The sense amplifier latches this stable state at a fixed sampling time, thereby forming a noisy initial response. The determination form for a single response bit can be expressed as:

[0089]

[0090] Among them, response bits Noisy initial response The Middle The read result of the bit, taking a value of 0 or 1, serves as the raw input for the server to subsequently perform temperature drift compensation and blind search comparison; indicator function : Logical decision operator, outputs 1 when the condition within parentheses is true, otherwise outputs 0. Its function is to convert the analog node voltage difference into a binary response; internal node voltage With internal node voltage They represent the first Each storage unit at the sampling time The voltages at both nodes fall within the array supply voltage window; sampling time The time when the sensing amplifier latches is set to a preset delay point after power-on, which serves to unify the latch boundaries of different authentication rounds.

[0091] In an embodiment of an implantable pacing terminal, the medical terminal receives a transient challenge vector. Then, the address decoder first selects the target word line, and the power-on controller pulls up the unit power supply according to a predetermined sequence. After the sensing amplifier completes latching, it sets each response bit... Serial connection to form a noisy initial response In parallel implementations of bedside attached monitoring terminals, the physically unclonable function of the storage array can also employ a register file-type power-on unit, provided the response bits... The result is still derived from the metastable release, meaning it remains within the same underlying principle. When using it, the transient challenge vector... It directly enters the physical array after generation, reducing intermediate dwell stages and noisy initial response. It more faithfully reflects the physical release state at that time.

[0092] Obtaining the noisy initial response Subsequently, the medical terminal does not perform algebraic error correction, fuzzy extraction, or auxiliary data derivation locally, but instead processes the noisy initial response. Including body temperature (in plain text) Downgrade indicator This information is then encapsulated into the authentication report message. Since the server will later need to verify the plaintext body temperature data... A thermal drift compensation surface is established, and a reverse comparison is performed based on the finite substitution space of shuffling with equal weight positions. Then, the terminal side performs a noisy initial response. Performing local purification can actually leak additional structural information.

[0093] Therefore, the payload field of the reported message only includes the terminal identifier and the noisy initial response. Body temperature readings Downgrade indicator The session sequence number and link verification fields should not have auxiliary data fields, error correction synthesis fields, or dynamic internal masks. Derived summary field.

[0094] In terms of the sequence of actions, the medical terminal first freezes the control registers used to drive the switching stage, and then performs a noisy initial response. Load into the send buffer, then clear the saved dynamic internal mask. The security register is then released, and finally the cross-switch network and challenge loader are released. If the wireless link returns a retransmission request, the medical terminal retransmits the cached same reporting message without regenerating the dynamic internal mask. Furthermore, it avoids rereading the physically unclonable function of the storage array, thus preventing multiple original samples available for external comparison within the same session. Link encapsulation can utilize the framing mechanism and message authentication field from the medical terminal communication stack; this step limits the message to not containing auxiliary data or dynamic internal masks. This action chain is cleared before it is sent.

[0095] As a supplement: the medical terminal will use transient challenge vectors The signal is sent to the word lines and bit lines of the physically unclonable function addressing chain of the memory array, and controlled power-on is performed after the target cell is selected; the sense amplifier operates at a fixed sampling time. The voltage difference between the two nodes of each memory cell is latched, and a noisy initial response is generated. Sampling time The timing characteristics of the sensing amplifier are predetermined during terminal registration and embedded in the security firmware. The authentication message fields of a medical terminal should at least include the terminal identifier, session number, and noisy initial response. Body temperature readings Downgrade indicator Link verification field, excluding other auxiliary data fields or error correction comprehensive fields.

[0096] In practice, the verification server obtains the necessary information to perform subsequent temperature drift compensation and blind search decisions, while the terminal side leaves no additional error correction evidence or mask remnants for reverse modeling. The same-weight shuffling process transforms the dynamic internal mask generated in step one... After being transformed into a local switching action of a cross-connect network, the baseline challenge vector The Hamming weights remain unchanged, and the server subsequently faces transient challenge results where the positions have been altered but the statistical skeleton has not been rewritten. Transient challenge vector. Noisy initial response after power-on readout of the direct-drive physical array It maintains tight coupling with the current metastable release process, reducing the additional observables caused by general caching and secondary processing.

[0097] The verification server will transmit body temperature in plaintext. Substituting the preset thermodynamic equations, the noisy initial response is... The corresponding digital twin baseline is temperature-oriented shifted to remove systematic thermal drift, resulting in the predicted response at the current body temperature. and shrinkage coefficient .

[0098] Step 2 Report Noisy Initial Response This includes both individual differences that can be used for authentication and systematic offsets caused by body temperature. These systematic offsets are not random noise, but rather caused by decreased carrier mobility, slight shifts in threshold voltage, and changes in the charging and discharging rates of cross-coupled nodes. If the verification server treats the reference responses in the registry as static templates, it can confuse body temperature-driven regular offsets with device-specific differences.

[0099] Therefore, step three involves first reconstructing the device's thermal state based on body temperature, then mapping the thermal state to a bit flip trend, and finally applying the trend to the single-chain processing of the digital twin baseline. The execution is primarily performed by the message parser, thermal equation engine, twin parameter library, and compensation solver within the verification server.

[0100] Verify server receives noisy initial response Body temperature readings and downgrade sign Then, the message parser first retrieves the reference temperature written during the registration of the medical terminal from the twin parameter database based on the terminal identifier. Reference migration rate Thermosensitive index This is related to the bit-level barrier parameter set. The reason for this approach is that even if different medical terminals use the same layout, their channel doping differentials, memory cell mismatch directions, and packaging thermal paths will differ. Without a device-specific parameter set, the body temperature data would be lost. It can only provide rough environmental information and cannot pinpoint specific response details.

[0101] The thermal equation engine then adjusts the reference mobility. With body temperature Calculate the current mobility coefficient:

[0102]

[0103] Among them, the current mobility coefficient Body temperature (in plain text) The equivalent carrier mobility under the corresponding thermal state is a real number greater than 0; the reference mobility... The medical terminal is at the reference temperature. The equivalent carrier mobility obtained from the registration is a real number greater than 0; body temperature plaintext The terminal local body temperature reported in step two is preferably measured using the Kelvin scale, with values ​​within the human body's suitable temperature range. Its function is to drive the thermal equation engine to the current certified temperature. It is obtained from the unencrypted body temperature measurement value uploaded in step two. Reference temperature. The reference temperature used to generate the digital twin baseline during the registration phase is preferably set to room temperature for registration, serving to unify the thermal state reference. It is obtained by setting the temperature during the registration phase and writing it into the digital twin parameter library. (Thermal sensitivity index) : The intensity of mobility decay with temperature, with a positive real number. Its function is to fold the comprehensive differences in device material, channel structure and packaging thermal path into a single thermistor parameter. It is obtained by fitting multiple temperature points during the registration stage.

[0104] In one embodiment, the verification server is located on the hospital authentication gateway, and the twin parameter library stores the reference temperature used for each medical terminal registration. and thermal sensitivity index The implanted heart rhythm monitor sends the body temperature in plaintext to the message parser. The message parser first reads the plaintext body temperature data. Then, the thermal equation engine is used to calculate the current mobility coefficient. The plaintext body temperature is converted into the current mobility coefficient that can directly drive the compensation chain. After obtaining the current mobility coefficient Then, the compensation solver renders the bit flipping trend of each response bit at the current body temperature. Instead of using the same threshold correction, it uses the barrier difference, thermal coupling gain, and relaxation width written at the time of registration for each bit as a probability map according to the position and size of each response bit.

[0105] Because the mismatch directions of different cells in the physically unclonable function of a storage array are different, applying the same temperature offset to all response bits not only flattens individual characteristics but may also miss local mutation bits. The bit flip probability of the first response bit at the current body temperature is:

[0106]

[0107] Wherein, bit flip probability : No. Each response bit is in the body temperature plaintext The probability of a flip occurring under thermal conditions, with a value between 0 and one; potential barrier difference. : No. Each storage cell at the reference temperature The equivalent energy barrier difference between the two steady states is a real number and its function is to describe which steady state the site naturally tends to be; it is obtained by fitting multiple temperature point repeated readouts during the registration phase.

[0108] Thermal coupling gain : No. The sensitivity of each memory cell to mobility changes, taking a real value, serves to translate temperature-induced device velocity changes into a barrier correction value; obtained from the registration phase fitting; relaxation width. : No. The transition width of each storage cell when it collapses from a metastable state to a steady state is a real number with a value greater than 0, and its function is to control the slope of the probability mapping.

[0109] Current mobility coefficient Compared with reference mobility The meaning is the same as before; the ratio of the two is used to characterize the relative thermal state shift. Exponential function It is a natural index mapping, and its function is to compress the energy barrier correction result into a comparable probability range.

[0110] Bit Flip Probability After the result is obtained, the compensation solver applies it to the reference sign bit saved during the registration phase. Above, generate the predicted sign bit at the current body temperature. :

[0111]

[0112] Among them, the predicted sign bit : No. Each response bit is a prediction result obtained after thermal drift stripping at the current body temperature, with a value of 0 or 1, and its function is to form the predicted response. ; Reference sign bit : No. Each response bit represents the reference state in the digital twin baseline during the registration phase, taking a value of 0 or 1, and its function is to provide the baseline orientation before translation; bit flip threshold. : No. The decision threshold for switching a response bit from the original baseline state to the baseline flip state, with a value between 0 and 1; bit flip probability. The meaning is the same as before; its function is to determine whether the position should be shifted at the current body temperature.

[0113] In a representative implementation, the twin parameter library stores the reference symbol bits for each medical terminal. Barrier difference Thermal coupling gain Relaxation width Bit Flip Threshold The compensation solver calculates the bit-flip probability bit by bit in the order of the response bits. To form a predictive response If the sampling points obtained during registration are sparse temperature regions, the compensation solver first performs piecewise conformal interpolation, and then substitutes them into the probability mapping.

[0114] When used, the verification server obtained a predicted response that matched the current body temperature. This allows us to separate systemic thermal drift from identity differences.

[0115] Predicted response After generation, the verification server also needs to provide step four with a quantized output that can directly drive threshold tightening. If the common statistical variance expression is used, although it can describe the degree of dispersion, it cannot uniformly incorporate bit-level confidence, differences before and after temperature drift compensation, and important bit weights in the registration stage.

[0116] Therefore, the shrinkage coefficient is generated in the form of residual misalignment mass ratio. The verification server simultaneously stores the bit weight of each response bit in the twin parameter library. and reference prediction position Then, the currently certified prediction sign bit Noisy initial response The results were compared and normalized to the misalignment at the reference temperature.

[0117]

[0118] Among them, the shrinkage coefficient : The mass ratio of the residual misalignment after current temperature compensation to the uncompensated misalignment at the reference temperature, taking a non-negative real number; Total number of response bits Noisy initial response The total number of bits, taking a positive integer value, is used to specify the range for summation. Bit weight. : No. Each response bit is weighted during the registration phase based on stability and discriminative power, and its value is a real number greater than 0. Its purpose is to prevent unstable weak bits from negatively impacting the contraction coefficient. Excessive impact; current response bit Noisy initial response The Middle The actual read value of a bit is either 0 or 1.

[0119] Predict sign bit The meaning is the same as before; its function is to provide a compensated baseline for comparison at the current body temperature; reference prediction position. At reference temperature The baseline prediction bit used for comparison, with a value of 0 or 1, serves to provide a reference misalignment object needed for normalizing the denominator; the protection constant. : To prevent small positive numbers with a denominator of zero, the value is taken as a positive real number; its purpose is to maintain calculation stability; absolute value sign The magnitude mapping of binary position difference is used to record the same value as zero and different values ​​as 1.

[0120] In terms of specific implementation, the server was verified to have a shrinkage coefficient. After calculation, the predicted response will be... and shrinkage coefficient Write to the session result cache if a downgrade flag is present. If invalid, step four directly reads the predicted response. and shrinkage coefficient If the downgrade flag is displayed For effectiveness, step four no longer performs the regular branch comparison; only the predicted response is retained. and shrinkage coefficient , as session control information.

[0121] When using this method, step four does not require reassessing the temperature drift peeling quality; it can be directly based on the shrinkage coefficient. This directly determines the direction and extent of tightening the tolerance threshold. (Broadcast body temperature) Converted to the current mobility coefficients via the thermal equation engine Subsequently, the verification server obtains a thermal state representation consistent with the device material and packaging thermal path, and the bit flip probability. Rendered bit by bit and used for reference sign bits Then, predict the response. Systematic thermal drift can be reduced from the noisy initial response. Peel it apart.

[0122] Step 4: Verify the server based on the contraction coefficient. The decision tolerance threshold is tightened in reverse, and the noisy initial response is processed within the finite reverse deshuffling space corresponding to the same-weight shuffling. Perform concurrent blind search comparison; when the regular path is not applicable, then use a degradation flag. Trigger the isolation verification path.

[0123] Step two is to hide the dynamic internal mask. No auxiliary data was transmitted to the verification server, therefore the server could not directly know the noisy initial response. Which shuffle branch does this correspond to? Although step three already uses body temperature in plaintext... By removing the systematic thermal drift, the predicted response was obtained. and shrinkage coefficient However, if the judgment is still based on the room temperature tolerance threshold during the registration stage, the branch expansion caused by the limited reverse unmixing will amplify the risk of misidentification.

[0124] Therefore, step four does not use blind search alone, but instead uses the shrinkage coefficient. This causes threshold tightening and finite reverse unmixing and concurrent screening to couple into a continuous chain, making branch expansion and threshold compression occur simultaneously.

[0125] Before entering a branch search, the verification server first reads the reference tolerance threshold saved during the registration phase by the threshold controller. Then, based on the shrinkage coefficient output in step three... Calculate the tightening threshold for the current session The shrinkage coefficient is used here. As a threshold adjustment factor, instead of directly using body temperature as the explicit value. The regulation is due to the explicit body temperature readings. It only reflects the thermal environment and cannot directly indicate the degree of residual misalignment after thermal drift is stripped away; the coefficient of shrinkage The thermal equation compensation result from step three has now been folded into a single quantity that can be directly used for decision-making. Tighten the threshold. The calculation form is:

[0126]

[0127] Among them, tightening threshold The tolerance limits used in this certification for decision-making are non-negative integers, and their function is to constrain the acceptable misalignment amount of subsequent reverse unshuffling branches; the reference tolerance threshold... The basic tolerance limits stored at the reference temperature during the registration phase are non-negative integers; the shrinkage coefficient. The residual misalignment quality ratio output in step three is a non-negative real number, representing the degree of shrinkage of the residual misalignment relative to the reference misalignment after temperature drift stripping; it also tightens the gain. Threshold controller for contraction coefficient The response intensity, taking a positive real number, controls the tightening speed; the floor operator. Rounding down brings the results of consecutive calculations down to an integer threshold that can be directly used by the comparator.

[0128] In one embodiment, after receiving the authentication message from the implanted heart rhythm monitor, the verification server first retrieves the contraction coefficient from the session result cache. The threshold controller uses this to set the reference tolerance threshold. Compression to tightening threshold The threshold will then be tightened. Write to the local registers of all branch comparators. If the downgrade flag is... For it to be effective, the threshold controller still completes the tightening of the threshold. The calculation is performed, but the value is kept in the session context and not sent to the regular branch array.

[0129] When using it, the shrinkage coefficient obtained in step three The decision boundary is directly transformed into an executable decision boundary in step four, and the regular path and the isolated path share the same prior session state without being confused with each other.

[0130] After setting the threshold, the verification server uses the branch scheduler to expand a finite set of reverse deshuffling corresponding to the same-weight shuffling structure in step two. Here, we do not enumerate the entire permutation space, but only the set of reverse exchange sequences that the crossbar switch network can generate under a given number of levels and a given control bit width. The total number of branches is denoted as... .

[0131] In step two, the shuffling of bits with the same weight is limited by the number of swap levels, the length of the control register, and the adjacent swapping rules. The server only needs to reproduce this space and does not need to traverse the permutation region that cannot be generated by the terminal hardware. For the second step... The reverse unmixing branch is used, and the branch comparator first processes the noisy initial response. After the reverse permutation operator Restore to the candidate order and then compare with the predicted response By comparing each bit, the branch misalignment amount is obtained:

[0132]

[0133] Among them, branch misalignment : No. Reverse unmixing branch and predicted response The total weighted misalignment between them, taking the value of a non-negative real number, serves as the basis for branch sorting; the total number of response bits. Noisy initial response The total number of digits, which takes the value of a positive integer, is used to specify the range for digit-by-digit comparison.

[0134] Bit weight Step 3 inherits the first The bit weight, which takes a real number greater than 0, is used to suppress the interference of weakly stable bits on branch sorting; it is obtained by normalizing the consistency ratio of repeated reads from each temperature node during the registration phase.

[0135] Inverse permutation component Noisy initial response In the After the reverse unmixing branch action, the first The candidate value for the bit is either 0 or 1; predict the sign bit. Predicting the response The Middle The predicted value of the bit, taking a value of 0 or 1, serves to provide a baseline for comparison after temperature drift stripping; absolute value sign. Binary difference mapping is used to assign 0 to consistency and 1 to inconsistency. (Branch numbering) : Reverse unshuffling branch index, with a value that is a positive integer and not greater than the total number of branches. Its function is to distinguish different candidate paths.

[0136] The branch comparator obtains the total branch misalignment. Then, the priority queue controller retrieves the minimum branch misalignment. Misalignment with the second smallest branch Then, the branch separation degree is calculated by the uniqueness discriminator:

[0137]

[0138] Among them, branch separation : The degree of separation between the best branch and the second-best branch, taking a non-negative real number, used to determine whether the optimal branch is sufficiently unique; Minimum branch misalignment. The minimum value among all branch misalignments, which is a non-negative real number, is used to indicate the error level of the optimal matching branch.

[0139] Secondary branch misalignment The second smallest value among all branch misalignments, taking the value of a non-negative real number and not less than the smallest branch misalignment. Its function is to serve as a unique comparison benchmark; the constant 1 is used to avoid the minimum branch misalignment. When the denominator is too small when it is close to zero, it leads to branch separation. Instability.

[0140] In a representative implementation, the verification server pre-writes the reverse deshuffling branches as read-only routing tables. The branch scheduler retrieves a set of routing tables at a time and sends them to the concurrent comparison array. Each branch array contains a reverse switching unit chain, a weighted summation tree, and a local minimum register. The minimum branch misalignment of the optimal branch... Not higher than the tightening threshold and branch separation If the uniqueness exceeds the preset threshold, the best branch is confirmed as the authentication hit branch.

[0141] When using it, the dynamic internal mask hidden in step two Without requiring recovery, the server can still filter candidate paths within the limited hardware reach and use branch separation. Eliminate ambiguity caused by multiple interpretations.

[0142] Inverse permutation operator Indicates the relationship with the first The positional restoration rules corresponding to each candidate shuffle path are used to reverse the noisy initial response after equal-weighted positional shuffling, restoring it to the original arrangement order under that candidate path. Reverse permutation operator. Only the order of the response bits is changed; the values ​​of the response bits remain the same. The verification server sequentially inputs the noisy initial response into different inverse permutation operators. Multiple candidate restoration results are obtained, and then each candidate restoration result is compared with the predicted response to determine the candidate path that best matches the current authentication.

[0143] When the downgrade sign If the hit branch is obtained through an invalid and normal path, the server output verifies the decision and records the hit branch number and the minimum branch misalignment. and tightening threshold Write to the session archive area, and then send a pass message back to the medical terminal. If a downgrade flag is present... For effective, or although downgraded flag Invalid but the optimal branch does not meet the tightening threshold If a unique boundary is reached, the verification server will no longer expand the regular branches but will switch to the isolated verification path. The execution entities of the isolated verification path include the verification server, the isolation router in the medical terminal, the static mask area, and the fuse driver. The server first sends an isolation command frame, which contains the session sequence number, the isolation route number, and a one-time acknowledgment bit. Upon receiving the frame, the medical terminal immediately shuts down the word line driver and sense amplifier power of the main array, causing the physical non-cloning function of the regular storage array to exit the current round of authentication. Then, the isolation router switches the authentication addressing to the pre-recorded static mask area. The static mask area stores only a small number of pre-registered isolation patterns, which correspond one-to-one with independent fuse units. The medical terminal reads the corresponding isolation pattern according to the isolation route number, forms an emergency verification message, and sends it back to the verification server. After the verification server compares the pattern with the isolation registry, it sends back a fuse confirmation command. The fuse driver then applies a fuse pulse to the corresponding fuse unit, invalidating the recently used isolation pattern.

[0144] In one embodiment, the bedside attached monitoring terminal is subject to interference from a strong electromagnetic environment, and a degradation flag has already been set in step one. For effectiveness, after the verification server sends the message, it directly skips the branch filtering and writes the isolation route number into the isolation command frame. The terminal's isolation router will disconnect from the main array and switch to the static mask area. After the emergency verification is completed, the fuse driver will blow the fuse unit once, and the isolation pattern will no longer be readable.

[0145] As a supplement: the medical terminal has a static mask area independent of the main array, which has multiple pre-registered isolation patterns. With multiple fuse units One-to-one correspondence, if downgrade flag It is valid; the verification server sends the isolation route number. The medical terminal shuts down main array communication and switches to the static mask area to read the first... An isolation pattern This serves as an emergency verification message. After the verification server passes the comparison, it sends a fuse confirmation command to the medical terminal; the fuse driver in the medical terminal then sends a signal to the corresponding fuse unit. Send a fuse pulse to make the first An isolation pattern It will become invalid in subsequent authentication sessions.

[0146] In use, even if a normal path is forced to exit during life support scenarios, the system still retains one controlled isolation verification opportunity, while simultaneously preventing repeated exploration of the same path through physical circuit breaking. (Tighten thresholds) From the coefficient of shrinkage After derivation, the temperature drift stripping result in step three is converted into a directly executable threshold control action. Branch misalignment in the finite inverse unmixing space. Branch separation After combining, the verification server can verify the dynamic internal mask without knowing it. Under the premise of filtering out the only hitting branch, the demotion flag is displayed. When the triggered isolation verification path is combined with the fuse-driven action, the regular authentication link and the emergency authentication link are clearly separated within the same session, and the isolation pattern will not be reused.

[0147] As a supplement: During the registration phase, the verification server controls the medical terminals to maintain a reference temperature. And multiple calibration temperature nodes, repeatedly perform power-on readouts on the physically unclonable function of the storage array to form a multi-temperature response sample set; the verification server uses the majority vote result of each response bit at the reference temperature as the reference sign bit. Based on the flipping frequency of each response point at multiple temperature points, a nonlinear fitting method with boundary constraints is used to obtain the reference mobility. Thermosensitive index Level barrier difference Thermal coupling gain Relaxation width Bit flip threshold Simultaneously, the verification server determines the bit weight based on the consistency ratio of each response bit in repeated readouts at multiple temperature points. and reference tolerance threshold Uniqueness threshold Tighten gain and the unique reverse unmixing and shuffling set It is also written into the authentication parameter library for use in subsequent authentication stages.

[0148] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0149] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0150] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0152] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A trusted authentication method for medical terminals based on physically unclonable functions, characterized in that: include, After receiving the benchmark challenge vector, the medical terminal synchronously collects physiological electrical signals and body temperature plaintext, extracts thermal shot noise from the least significant bit of the continuous sampling window to generate a dynamic internal mask and outputs a degradation flag. The medical terminal uses a dynamic internal mask to drive a cross-switch network to perform same-weighted shuffling on the baseline challenge vector, generates a transient challenge vector and injects it into the SRAM-PUF, obtains the noisy initial response and then reports the noisy initial response, plaintext body temperature and degradation flag. The verification server performs thermal compensation on the digital twin baseline based on the plaintext body temperature to obtain the predicted response and shrinkage coefficient. It then tightens the tolerance threshold based on the shrinkage coefficient and performs blind search comparison on the noisy initial response within a limited reverse demixing space. Alternatively, it switches to the static mask isolation area to perform physical fuse verification when the degradation flag is valid.

2. The trusted authentication method for medical terminals according to claim 1, characterized in that: After receiving the benchmark challenge vector, the medical terminal drives the physiological electrical signal sampling channel and the body temperature sampling channel with the same hardware clock. Within the same authentication cycle, it obtains the corresponding sampling sequence and body temperature plaintext, and extracts the least significant bit thermal shot noise within the continuous sampling window to generate the dynamic internal mask and simultaneously determine the degradation flag.

3. The trusted authentication method for medical terminals according to claim 2, characterized in that: The medical terminal determines the starting offset of the continuous sampling window according to the benchmark challenge vector, extracts binary results from the target bit plane, and generates a dynamic internal mask according to the predetermined extraction step size and folding order. The total length of the dynamic internal mask is consistent with the total number of control bits of the cross switch network. After generation, it resides in the local hardware only during this round of certification.

4. The trusted authentication method for medical terminals according to claim 3, characterized in that: The medical terminal writes the dynamic internal mask into the inter-stage control register of the cross-connect network, and each level of the switching unit performs in-situ pass-through or position swapping on adjacent challenge bits in the reference challenge vector, outputting a transient challenge vector. The same-weight bit shuffling only changes the bit position arrangement of the reference challenge vector, without changing the Hamming weight of the reference challenge vector.

5. The trusted authentication method for medical terminals according to claim 4, characterized in that: The medical terminal sends the transient challenge vector into the word line and bit line addressing chain of the SRAM-PUF. After the target cell is selected, it performs controlled power-on and latches the internal node voltage difference of each memory cell at a fixed sampling time to form a noisy initial response. The authentication messages reported by the medical terminal include a noisy initial response, plaintext body temperature, and a downgrade flag, but do not include auxiliary data.

6. The trusted authentication method for medical terminals according to claim 5, characterized in that: During the registration phase, the verification server establishes a digital twin parameter library for each medical terminal, storing reference temperature, reference mobility, thermal index, bit-level barrier difference, thermal coupling gain, transition width, bit flip threshold, and reference sign bit. During the authentication phase, the plaintext body temperature is substituted into the thermodynamic equation to obtain the bit flip trend under the current thermal state, and based on this, the digital twin baseline is thermally shifted to generate a predicted response.

7. The trusted authentication method for medical terminals according to claim 6, characterized in that: The verification server calculates the shrinkage coefficient, which characterizes the degree of residual misalignment after temperature drift stripping, based on the bit-level misalignment relationship between the noisy initial response and the predicted response, and in combination with the bit weights determined for each response bit during the registration phase. The verification server stores the shrinkage coefficient in correspondence with the uncompensated misalignment result under the reference state, as input for subsequent tightening tolerance threshold.

8. The trusted authentication method for medical terminals according to claim 7, characterized in that: The verification server reads the reference tolerance threshold saved during the registration phase and generates the tightening tolerance threshold for the current session based on the shrinkage coefficient. Then, in the finite reverse deshuffling space corresponding to the same weight shuffling, the noisy initial response is compared concurrently, the misalignment of each candidate branch relative to the predicted response is calculated, and the degree of separation between the smallest misaligned branch and the second smallest misaligned branch is used to determine whether there is a unique hit branch.

9. The trusted authentication method for medical terminals according to claim 8, characterized in that: When the downgrade flag is valid, the verification server stops the regular authentication path and sends an isolation routing command to the medical terminal; the medical terminal shuts down the main array communication and switches to the static mask isolation zone, reads the isolation pattern corresponding to the isolation routing command, forms an emergency verification message and returns it to the verification server.

10. The trusted authentication method for medical terminals according to claim 9, characterized in that: The static mask isolation zone contains multiple pre-registered isolation patterns, and each isolation pattern is associated with a corresponding physical fuse unit. When the verification server confirms that the emergency verification message is consistent with the isolation registration database, the medical terminal applies a fuse pulse to the physical fuse unit called for this authentication, causing the corresponding isolation pattern to become invalid in subsequent authentication sessions.

Images