Data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform

Abstract

This invention discloses a smart cloud-based electronic scale platform for data encryption and anti-cheating, belonging to the field of smart weighing supervision technology. It includes an encryption processing module that acquires the original weight signal and encrypts the data using a symmetric algorithm to dynamically generate a session key, and an asymmetric algorithm to encrypt the session key to generate an encrypted weight data packet. A synchronous upload module uploads the data packet, device identification code, and timestamped logs to the cloud via an encrypted link. A cloud decryption and restoration module uses a paired private key to decrypt the session key, symmetrically decrypting the basic weight sequence. A weight calibration module combines a benchmark sensor calibration curve with a dynamic environmental compensation model to calibrate and generate a standard weight value. An audit storage module associates and binds the standard value, logs, and device code to an anti-tampering audit library. This platform achieves dynamic encryption to prevent cheating during a single weighing, and cloud-based dynamic compensation calibration to improve accuracy, forming an undeniable weighing record.

Description

Technical Field

[0001] This invention belongs to the field of intelligent weighing supervision technology, specifically a data-encrypted anti-cheating intelligent cloud weighing supervision electronic scale platform. Background Technology

[0002] Existing electronic scales mostly employ conventional technical solutions for data encryption and weight calibration. For encryption, they commonly use symmetric encryption algorithms with fixed keys, or directly use asymmetric encryption algorithms to encrypt weight data; for calibration, they rely on factory-stored static sensor calibration curves to correct sensor signals according to fixed parameters.

[0003] Existing technical solutions have shortcomings: long-term use of fixed keys is prone to leakage, posing a risk of cracking all historical and future weighing data; single asymmetric encryption is computationally complex and difficult to adapt to high-frequency weighing scenarios. Weight calibration is based solely on static calibration curves and does not consider the continuous interference of dynamic environmental changes such as temperature, humidity, and vibration on sensor signals during actual use, leading to the gradual accumulation of measurement deviations in complex scenarios and making it difficult to guarantee long-term accuracy.

[0004] The core problems this invention addresses are: for the encryption stage, how to generate independent dynamic security protection for each weighing operation to avoid the risks of fixed keys while maintaining efficiency; and for the calibration stage, how to overcome the limitations of static calibration in cloud processing and introduce dynamic environmental factors to correct weight data, thereby improving calibration accuracy. Through these two improvements, a tamper-proof and highly reliable weighing supervision foundation is formed, meeting the dual requirements of data fraud prevention and measurement accuracy. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art;

[0006] To this end, the present invention proposes a data encryption and anti-cheating intelligent cloud weighing supervision electronic scale platform, comprising:

[0007] The encryption processing module is used to acquire the raw weight signal data generated by each weighing operation of the electronic scale, and generate an encrypted weight data packet after conversion and processing. The symmetric encryption algorithm uses a dynamically generated session key, and the asymmetric encryption algorithm is used to encrypt the session key.

[0008] The synchronous upload module is used to synchronously upload the encrypted weight data packet, the unique device identification code of the electronic scale, and the weighing operation log containing timestamps to the cloud monitoring server through a preset encrypted communication link.

[0009] The decryption and restoration module, on the cloud monitoring server, uses an asymmetric private key paired with the electronic scale to decrypt the session key in the encrypted weight data packet, and uses the decrypted session key to symmetrically decrypt the weight data in the encrypted weight data packet to restore the decrypted basic weight digital sequence.

[0010] The weight calibration module is used to perform calibration and correction processing on the decrypted basic weight digital sequence in the cloud based on the pre-stored benchmark sensor calibration curve and dynamic environment compensation model, and generate the calibrated standard weight value.

[0011] The audit storage module is used to associate and bind the calibrated standard weight value with the weighing operation log and the device identification code, and store them in the cloud-based anti-tampering audit database to form an irrefutable weighing transaction record.

[0012] Further, the process of generating an encrypted weight data packet after conversion includes:

[0013] The system acquires the raw weight signal data generated by each weighing operation of the electronic scale, and performs analog-to-digital conversion and noise reduction filtering on the raw weight signal data to generate a basic weight digital sequence.

[0014] The basic weight digital sequence is encrypted by combining a symmetric encryption algorithm with an asymmetric encryption algorithm to generate an encrypted weight data packet. The symmetric encryption algorithm uses a dynamically generated session key, and the asymmetric encryption algorithm is used to encrypt the session key.

[0015] The step of performing analog-to-digital conversion and noise reduction filtering on the original weight signal data to generate a basic weight digital sequence includes:

[0016] The analog voltage signal output by the electronic scale sensor is sampled and quantized by a high-precision analog-to-digital converter to generate an initial weight digital sequence.

[0017] For the initial weight digital sequence, an adaptive sliding window midpoint filtering algorithm is used to remove abnormal jump points introduced by mechanical vibration or transient electromagnetic interference;

[0018] The filtered sequence is further processed by multi-scale decomposition based on wavelet transform to separate the high-frequency noise component and the low-frequency trend component of the signal.

[0019] The high-frequency noise component is subjected to threshold processing to suppress random noise, and the processed high-frequency noise component is reconstructed with the low-frequency trend component to obtain a smoothed weight signal sequence.

[0020] The mean of a stable region is extracted from the smoothed weight signal sequence. The stable region is determined by a preset sequence variance threshold. The mean is the base weight digital sequence.

[0021] Furthermore, the step of performing hybrid encryption processing on the basic weight digital sequence using a combination of symmetric and asymmetric encryption algorithms to generate an encrypted weight data packet includes:

[0022] A random number seed is generated locally on the electronic scale, and the random number seed is enhanced using the physical random source of the hardware security module;

[0023] Based on the enhanced random number seed, a symmetric encryption session key for this weighing session is generated through a key derivation function;

[0024] The basic weight digital sequence is encrypted using the Advanced Encryption Standard algorithm and the symmetric encryption session key to generate symmetric encrypted ciphertext.

[0025] Obtain the public key from the asymmetric key pair pre-set on the cloud monitoring server, and use the public key to perform asymmetric encryption on the symmetric encryption session key to generate the encrypted session key;

[0026] The symmetric encrypted ciphertext, the encrypted session key, and the message authentication code used for integrity verification are encapsulated and packaged to form the encrypted weight data packet.

[0027] Furthermore, the calibration and correction process performed on the decrypted basic weight digital sequence in the cloud based on pre-stored benchmark sensor calibration curves and dynamic environmental compensation models, generating a calibrated standard weight value, includes:

[0028] Based on the unique device identification code of the electronic scale, the corresponding benchmark sensor calibration curve is retrieved from the calibration database stored in the cloud. The benchmark sensor calibration curve describes the mapping relationship between the sensor's original digital signal and the actual physical weight.

[0029] Extract the temperature and humidity environmental parameters at the moment the weighing occurred from the weighing operation log;

[0030] The environmental parameters are input into the dynamic environmental compensation model, which is constructed based on the influence of temperature and humidity on sensor sensitivity and zero-point drift.

[0031] Based on the output of the dynamic environment compensation model, the sensor characteristic correction coefficient for the current environment is calculated.

[0032] The decrypted base weight digital sequence is substituted into the reference sensor calibration curve to obtain a preliminary weight value. Then, the preliminary weight value is multiplied by the sensor characteristic correction coefficient to obtain the calibrated standard weight value.

[0033] Furthermore, it also includes a behavior detection module, used to detect weighing cheating behavior through multi-dimensional data analysis, as detailed below:

[0034] Extract the calibrated standard weight value sequence of the same electronic scale or the same batch of goods from the anti-tampering audit database in chronological order;

[0035] Statistical analysis was performed on the calibrated standard weight value sequence to calculate the mean and variance of the weight values, and a weight trend graph over time was constructed.

[0036] The calibrated standard weight value of the current weighing is compared with the average value of the historical sequence. If the deviation exceeds the preset reasonable fluctuation threshold, a weight abnormality event is triggered.

[0037] Simultaneously, the weighing operation log was analyzed to identify suspicious operation patterns during the weighing process, including abnormally short time intervals between the placement and removal of the weighing tray, and repeated weighing operations before the weighing stabilized.

[0038] Logically correlate the weight anomaly event with the suspicious operation mode. If they occur simultaneously, generate a cheating suspicion alarm and store the cheating suspicion alarm, related weighing records, and related evidence chain into a special monitoring database.

[0039] Furthermore, the identification of suspicious operating patterns during the weighing process includes:

[0040] The state change sequence of each weighing event is parsed from the weighing operation log, and the state change sequence includes the transition between zero point state, stable state and overload state.

[0041] Calculate the first time interval from when the weight touches the weighing pan until the weight reading reaches the steady state;

[0042] Calculate the second time interval during which the weight reading returns from a steady state to zero after the object leaves the weighing pan;

[0043] If the first time interval or the second time interval is less than the limit time threshold for device physical response and signal processing, it is determined to be an abnormality in rapid pallet placement and removal;

[0044] The system counts the number of times the weight reading switches between the stable state and the zero point state within a single settlement cycle. If the number of switching exceeds a preset frequent operation threshold, it is determined to be an abnormal repeated weighing operation.

[0045] Furthermore, it also includes: an evidence preservation enhancement module, used to enhance the immutability of weighing records based on blockchain technology;

[0046] The calibrated standard weight value, the weighing operation log, and the device identification code, which have been linked and bound together, are used to generate a structured weighing transaction record.

[0047] Calculate the hash value of the structured weighing transaction record;

[0048] The hash value is packaged together with the timestamp and the hash value of the previous block to form the data of a new block;

[0049] The new block is digitally signed using the private key of the cloud monitoring server;

[0050] The digitally signed new block is broadcast to a pre-built permissioned blockchain network for consensus verification. Once verified, the new blockchain is connected to a regulatory blockchain composed of data from multiple electronic scales.

[0051] Furthermore, the step of packaging the hash value together with the timestamp and the hash value of the preceding block to form a new block includes:

[0052] Obtain the block header hash value of the latest block on the blockchain, and use it as the hash value of the preceding block;

[0053] Obtain the current accurate time from a trusted system time source and use it as the authoritative timestamp for this weighing transaction;

[0054] Create a new data structure that contains at least a data body field, a timestamp field, and a previous hash field;

[0055] The hash value of the structured weighing transaction record is filled into the data body field, the authoritative timestamp is filled into the timestamp field, and the hash value of the preceding block is filled into the preceding hash field.

[0056] The new data structure is serialized and encoded to generate the original data packet of the new block.

[0057] Furthermore, it also includes: a firmware security module, used to implement secure firmware startup and key protection on the electronic scale:

[0058] After the electronic scale is powered on, its main controller reads the boot program code and the preset root certificate public key hash value from the read-only memory;

[0059] Read the signed firmware image and corresponding digital signature from the secure storage chip;

[0060] The validity of the digital signature is verified using the hash value of the root certificate public key to ensure that the firmware image has not been tampered with;

[0061] After successful verification, the firmware image is loaded and run;

[0062] During firmware operation, the device private key used for the asymmetric encryption algorithm, or the seed key used to generate the session key, is read from the physically isolated area of ​​the secure storage chip. All encryption and decryption operations are performed in a secure area within the chip, and the private key and seed key never leave the secure storage chip.

[0063] Furthermore, it also includes: an intelligent analysis module for self-evolutionary identification of cheating behavior patterns based on machine learning;

[0064] Continuously collect all confirmed cheating case data from the special monitoring database to form a cheating sample training set, and at the same time collect normal weighing data to form a normal sample training set.

[0065] Multidimensional feature vectors are extracted from the cheating sample training set and the normal sample training set. The multidimensional feature vectors include weight deviation features, operation timing features, environmental parameter features and device status features.

[0066] An ensemble learning classifier is trained using the multidimensional feature vectors, the ensemble learning classifier being a combination of decision trees, support vector machines and neural network base classifiers;

[0067] The trained ensemble learning classifier is deployed to a cloud-based data analytics engine for real-time analysis of newly uploaded weighing transaction records.

[0068] Based on the output of the ensemble learning classifier, a cheating risk probability score is calculated for newly occurring weighing behavior, and records with cheating risk probability scores higher than the dynamically adjusted threshold are automatically marked as high-risk records and pushed to the supervisor's interface for review.

[0069] Compared with the prior art, the beneficial effects of the present invention are:

[0070] A unique session key is dynamically generated for each weighing operation. The converted weight signal data is encrypted using a symmetric encryption algorithm to generate an encrypted weight data packet. Simultaneously, the session key is encrypted using an asymmetric encryption algorithm and encapsulated within the data packet. The dynamic session key ensures independent encryption of each weighing operation, meaning that even if a key is leaked in one operation, it will not affect subsequent operations. Asymmetric encryption protects the session key during transmission, preventing interception by intermediate nodes. This mechanism replaces fixed keys or single encryption methods, giving the encrypted weight data packet both individual independence and reliable key distribution, effectively blocking data tampering and cheating paths.

[0071] The cloud-based weight calibration module calls upon pre-stored benchmark sensor calibration curves and simultaneously incorporates a dynamic environmental compensation model to calibrate the decrypted base weight digital sequence. The dynamic environmental compensation model adjusts the offset and scaling factor of the calibration curve based on the environmental time period information implied in the timestamps of the weighing operation logs, offsetting the interference of dynamic factors such as temperature and humidity on the sensor signal. Compared to using only static calibration curves, this dual calibration method makes the calibrated standard weight value closer to the actual weight, reducing measurement deviations in complex environments and improving the long-term accuracy and reliability of weighing results. Attached Figure Description

[0072] Figure 1 This is a timing diagram of a data encryption and anti-cheating intelligent cloud weighing supervision electronic scale platform according to the present invention;

[0073] Figure 2 A flowchart for the conversion and processing of raw weight signal data;

[0074] Figure 3 A flowchart for cloud-based calibration and correction processing. Detailed Implementation

[0075] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0076] See Figure 1The platform architecture comprises components deployed locally on the electronic scale and components deployed on a cloud-based monitoring server, which collaborate via an encrypted communication link. On the electronic scale, the encryption processing module acquires the raw weight signal data generated by the sensor during each weighing operation. This data undergoes analog-to-digital conversion and filtering preprocessing to form a basic weight digital sequence. This module employs a hybrid encryption mechanism to process this sequence: first, a session key is dynamically generated, and this session key is used to encrypt the basic weight digital sequence using a symmetric encryption algorithm. Then, an asymmetric public key obtained from the cloud is used to encrypt the session key itself, ultimately packaging it into an encrypted weight data packet. The synchronization upload module is responsible for uploading this encrypted weight data packet, along with the electronic scale's unique device identifier and a timestamp log recording the operation action and time, to the cloud-based monitoring server in real time via a pre-defined encrypted communication link. On the cloud side, the decryption and restoration module uses the asymmetric private key paired with the electronic scale to decrypt the received data packet, restoring the session key. This session key is then used to symmetrically decrypt the weight ciphertext, obtaining the decrypted basic weight digital sequence. Subsequently, the weight calibration module retrieves the corresponding benchmark sensor calibration curve from the cloud database based on the device identification code. Combining this with real-time environmental parameters extracted from the logs, it calculates the environmental correction coefficient using a dynamic environmental compensation model. This corrects the decrypted base weight sequence, generating a high-precision calibrated standard weight value. The audit storage module precisely associates and binds this standard weight value with the corresponding weighing operation log and device identification code, storing it as a complete weighing transaction record in the cloud's tamper-proof audit database, forming an irrefutable and permanent record.

[0077] See Figure 2In one embodiment of the present invention, the encryption processing module generates encrypted weight data packets following the steps described above. The encryption processing module acquires the original weight signal data generated by each weighing operation of the electronic scale. The original weight signal data is a continuous analog voltage signal from a strain gauge load cell. The original weight signal data undergoes analog-to-digital conversion and noise reduction filtering to generate a basic weight digital sequence. Specifically, a high-precision analog-to-digital converter samples and quantizes the analog voltage signal at a sampling rate of one thousand times per second, generating an initial weight digital sequence containing one thousand sampling points. For the initial weight digital sequence, an adaptive sliding window midpoint filtering algorithm is used to remove abnormal jump points introduced by mechanical vibration or transient electromagnetic interference. In a specific implementation, the sliding window size can be set to ten sampling points. The filtered sequence is further processed using a multi-scale decomposition technique based on wavelet transform to separate the high-frequency noise component and the low-frequency trend component of the signal. Thresholding is applied to the high-frequency noise component to suppress random noise, and the processed high-frequency noise component and the low-frequency trend component are reconstructed to obtain a smoothed weight signal sequence. The reconstruction process is an inverse wavelet transform. The mean of the stable region is extracted from the smoothed weight signal sequence. The stable region is determined by a preset sequence variance threshold. In specific implementation, the variance threshold is set to be less than 0.001. The mean is the base weight digital sequence, which is a value representing the original digital reading of the sensor.

[0078] After generating the basic weight numerical sequence, a hybrid encryption process combining symmetric and asymmetric encryption algorithms is applied to the sequence to generate an encrypted weight data packet. A random number seed is generated locally on the electronic scale and enhanced using the physical random source of the hardware security module. Based on the enhanced random number seed, a symmetric encryption session key for this weighing session is generated using a key derivation function. In some embodiments, the symmetric encryption session key uses a 256-bit Advanced Encryption Standard (AES) key. The basic weight numerical sequence is encrypted using the AES algorithm and the symmetric encryption session key to generate symmetric encrypted ciphertext. The symmetric encryption process employs a cipher block chaining mode. A public key is obtained from a pre-configured asymmetric key pair on the cloud monitoring server. This public key is used to perform asymmetric encryption on the symmetric encryption session key to generate an encrypted session key. In some embodiments, the asymmetric encryption algorithm uses the RSA algorithm. Optionally, the encryption processing module calculates the hash value of the symmetric encrypted ciphertext and the encrypted session key to generate a message authentication code for integrity verification. It is understood that the message authentication code prevents data tampering during transmission. The symmetric encryption ciphertext, the encrypted session key, and the message authentication code are encapsulated and packaged to form an encrypted weight data packet. The structure of the encrypted weight data packet includes header information and the aforementioned data fields.

[0079] In some embodiments, the signal smoothing process employs a specific threshold function to handle high-frequency noise components. Optionally, the threshold function uses a soft thresholding rule, as shown in the formula: ,when Otherwise In the formula, the symbol Represents the high-frequency coefficients obtained from wavelet transform decomposition, with the symbol... Represents the wavelet coefficients after thresholding, with the symbol... Indicates the preset noise threshold, symbol This represents the sign function, the sign... This indicates the operation of taking the absolute value.

[0080] See Figure 3 In one embodiment of the present invention, the weight calibration module performs calibration in the cloud following the steps described above. The weight calibration module retrieves the corresponding reference sensor calibration curve from the calibration database stored in the cloud based on the unique device identifier of the electronic scale. The reference sensor calibration curve describes the mapping relationship between the sensor's original digital signal and the actual physical weight. In specific implementations, the unique device identifier is a unique serial number burned into each electronic scale at the factory. Temperature and humidity environmental parameters at the time of weighing are extracted from the weighing operation log received by the decryption and restoration module. In specific implementations, temperature and humidity environmental parameters are collected in real time by a temperature and humidity sensor integrated inside the electronic scale and written into the weighing operation log. The temperature and humidity environmental parameters are input into a dynamic environmental compensation model. The dynamic environmental compensation model is constructed based on the influence of temperature and humidity on sensor sensitivity and zero-point drift. It can be understood that the dynamic environmental compensation model is a mathematical model pre-fitted through calibration experiments. In some embodiments, the dynamic environmental compensation model uses temperature and humidity as input variables and outputs a comprehensive compensation factor for adjusting sensor readings. Based on the output of the dynamic environmental compensation model, a sensor characteristic correction coefficient for the current environmental conditions is calculated. This sensor characteristic correction coefficient is a multiplicative coefficient used to correct the weight. The decrypted base weight sequence is obtained from the decryption and restoration module. This base weight sequence represents the original sensor reading. The value of the decrypted base weight sequence is substituted into the benchmark sensor calibration curve to obtain a preliminary weight value. In practice, the benchmark sensor calibration curve is a lookup table or fitted polynomial that maps digital readings to kilograms. This preliminary weight value is multiplied by a sensor characteristic correction factor to finally generate the calibrated standard weight value, expressed in standard units.

[0081] In some embodiments, the reference sensor calibration curve is represented by a piecewise linear interpolation function, with each segment corresponding to the linear fitting result of the sensor within a specific measurement range. Optionally, the dynamic environmental compensation model includes compensation for sensor sensitivity temperature drift, which is approximately linearly related to temperature. In a specific implementation, the effect of temperature on zero-point drift is described using a second-order polynomial model. Temperature and humidity data extracted from the weighing operation log are used to calculate the sensor characteristic correction coefficient, which involves the synthesis of multiple drift factors. It can be understood that the sensor characteristic correction coefficient is used to correct the initial weight value to the weight under standard conditions. In some embodiments, the output of the dynamic environmental compensation model includes a multiplicative sub-term for the effect of humidity, which is more significant in high-humidity environments. Sensor characteristic correction coefficient The calculation formula is:

[0082]

[0083] Where: symbol This represents the final sensor characteristic correction factor, with the symbol... This represents the current temperature value extracted from the weighing operation log, symbol [symbol missing]. Represents the reference temperature value, symbol This represents the current humidity value extracted from the weighing operation log, symbol [symbol missing]. Indicates the reference humidity value, symbol Represents the temperature effect coefficient, symbol This represents the humidity influence coefficient. It can be understood as the reference temperature value. and reference humidity value These are the standard environmental conditions determined during sensor factory calibration. Optional, temperature influence coefficient. Humidity Influence Coefficient The data was obtained by fitting a large amount of calibration experimental data and stored in a cloud database, which is associated with the device identification code.

[0084] In one embodiment of the present invention, the behavior detection module detects the implementation process of weighing fraud through multi-dimensional data analysis, following the steps described above. The behavior detection module extracts a sequence of calibrated standard weight values ​​for the same electronic scale or the same batch of goods from an anti-tampering audit database in chronological order. Statistical analysis is performed on the calibrated standard weight value sequence to calculate the mean and variance of the calibrated standard weight values, and a weight trend graph showing the change of the calibrated standard weight values ​​over time is constructed. The current calibrated standard weight value is compared with the mean of the historical sequence. If the deviation exceeds a preset reasonable fluctuation threshold, a weight anomaly event is triggered. In specific implementations, the reasonable fluctuation threshold can be set to 5% of the historical average, depending on the type of goods.

[0085] In practical implementation, the behavior detection module simultaneously analyzes the weighing operation log to identify suspicious operation patterns during the weighing process. It parses the state change sequence of each weighing event from the log, including transitions between zero-point, stable, and overload states. The module calculates the first time interval from when the weight touches the weighing pan until the weight reading reaches a stable state; this first time interval is the time required for the weight signal to stabilize within a preset error band after the weight leaves the weighing pan. The module also calculates the second time interval from when the weight reading returns from a stable state to zero after the weight leaves the weighing pan; this second time interval is the time required for the weight signal to return from a stable value to near zero. If either the first or second time interval is less than the limit threshold for the device's physical response and signal processing, it is determined to be an abnormality in the rapid placement and removal of the weighing pan. In some embodiments, the limit threshold is set to 0.3 seconds. The module counts the number of times the weight reading switches between the stable and zero states within a single settlement cycle. If the number of switches exceeds a preset frequent operation threshold, it is determined to be an abnormality in repeated weighing operations; the frequent operation threshold can be set to 3 times. It is understandable that a single settlement cycle is defined as the period from the start of a zeroing operation to the completion of transaction confirmation.

[0086] The behavior detection module performs logical correlation analysis between abnormal weight events and suspicious operation patterns. If an abnormal weight event occurs simultaneously with abnormal rapid pallet placement and removal or repeated weighing operations, a cheating suspicion alarm is generated, and the alarm, relevant weighing records, and related evidence chains are stored in a special monitoring database. In some embodiments, the logical correlation analysis is a Boolean AND operation, meaning that a cheating suspicion alarm is generated when both the abnormal weight event and any suspicious operation pattern flag are true.

[0087] In practice, the intelligent analysis module continuously operates on a machine learning-based self-evolutionary identification process for cheating behavior patterns. The module continuously collects all confirmed cheating cases from a special monitoring database to form a cheating sample training set, while simultaneously collecting manually verified normal weighing data from an anti-tampering audit database to form a normal sample training set. Multi-dimensional feature vectors are extracted from both the cheating and normal sample training sets. These vectors include weight deviation features, operation timing features, environmental parameter features, and equipment status features. The operation timing features include a first time interval and a second time interval. An ensemble learning classifier is trained using these multi-dimensional feature vectors. This classifier combines decision trees, support vector machines, and neural network base classifiers; in some embodiments, gradient boosting is used as the ensemble strategy. The outputs of the base classifiers are combined through weighted voting or a meta-learner. The trained ensemble learning classifier is then deployed to a cloud-based data analysis engine for real-time analysis of newly uploaded weighing transaction records. Based on the output of the ensemble learning classifier, a cheating risk probability score is calculated for newly occurring weighing behavior. Records with cheating risk probability scores higher than the dynamic adjustment threshold are automatically marked as high-risk records and pushed to the supervisor's interface for review. The dynamic adjustment threshold is set by the administrator based on the historical false alarm rate.

[0088] Optionally, training the ensemble learning classifier involves a feature scaling step. The ensemble learning classifier processes the multidimensional feature vector and outputs a cheating risk probability score ranging from 0 to 1. Cheating risk probability score The calculation can be expressed as:

[0089]

[0090] Where: symbol This represents the final cheating risk probability score, with the symbol... Represents the input multidimensional feature vector, symbol Indicates the first The output function of each base classifier, symbol Indicates the first The weights of each base classifier in the ensemble, denoted by _____. Represents the total number of base classifiers, symbol This represents a normalized function that maps linear scores to the interval [0,1], such as a logistic function. (Cheating risk probability scoring) The closer the value is to 1, the higher the risk of cheating. This is understandable; the standardization function... Ensure that the cheating risk probability score has a probabilistic interpretation. Weights of the ensemble learning classifier. The algorithm is determined during the training process.

[0091] In one embodiment of the present invention, the implementation process of the evidence enhancement module to enhance the immutability of weighing records based on blockchain technology follows the steps described above. The evidence enhancement module receives output from the audit storage module, which includes the calibrated standard weight value after association and binding, the weighing operation log, and the device identification code. The evidence enhancement module formats and encapsulates this data to generate a structured weighing transaction record. The hash value of the structured weighing transaction record is calculated; in specific implementations, the hash calculation uses the SHA-256 algorithm. The hash value, along with the timestamp and the hash value of the preceding block, is packaged together to form a new block of data. The specific packaging process follows the defined data structure. The block header hash value of the latest block on the blockchain is obtained as the hash value of the preceding block. The current accurate time is obtained from a trusted time source in the system as the authoritative timestamp for this weighing transaction; in some embodiments, the trusted time source uses the network time protocol service provided by the National Time Service Center. A new data structure is created, which at least includes a data body field, a timestamp field, and a preceding hash field. The hash value of the structured weighing transaction record is filled into the data body field. The authoritative timestamp is entered into the timestamp field. The hash value of the preceding block is entered into the preceding hash field. The new data structure is serialized and encoded to generate the original data packet of the new block. In some embodiments, the serialization and encoding uses JSON format. In some embodiments, the original data packet of the new block includes an additional version number field to identify the blockchain protocol version. Optionally, the detailed content of the structured weighing transaction record can be recorded in a table, see Table 1:

[0092] Table 1: Fields of Structured Weighing Transaction Records

[0093]

[0094] As can be understood, the fields shown in Table 1 constitute a structured weighing transaction record, and the hash value of this record will be filled into the data body fields. Calculate the hash value of the structured weighing transaction record. Using a specific algorithm, the process can be summarized as follows:

[0095]

[0096] Where: symbol The hash value obtained after final calculation is represented by the symbol. Indicates the cryptographic hash function used, symbol This indicates that the structured weighing transaction records will be used. The operation of concatenating all fields into a single byte stream in a predetermined order. (Structured weighing transaction record) This includes specific information such as the device identification code, the calibrated standard weight value, and the weighing operation log. This can be understood as a hash function. It has unidirectional and impact resistance.

[0097] After generating the original data packet for the new block, the evidence enhancement module digitally signs the new block using the private key of the cloud-based monitoring server. The digitally signed new block is then broadcast to a pre-built permissioned blockchain network for consensus verification. In some embodiments, the permissioned blockchain network employs a practical Byzantine fault-tolerant consensus algorithm. Once verified, the verification nodes in the permissioned blockchain network connect the new block to the monitoring blockchain, which consists of data from multiple electronic scales, forming an immutable chain structure.

[0098] In one embodiment of the present invention, the firmware security module implements the secure boot and key protection process for the electronic scale firmware following the steps described above. After the electronic scale is powered on, its main controller reads the boot program code and the preset root certificate public key hash value from the read-only memory. In a specific implementation, the root certificate public key hash value is the SHA-256 digest of the manufacturer's public key, which is stored in the read-only memory. The signed firmware image and the corresponding digital signature are read from the secure storage chip. The signed firmware image contains all the application code required for the electronic scale to operate. The validity of the digital signature is verified using the root certificate public key hash value. The verification process involves decrypting and comparing the digital signature using the public key corresponding to the root certificate public key hash value. In a specific implementation, the digital signature is generated using an elliptic curve-based digital signature algorithm. After successful verification, the signed firmware image is loaded and run. The firmware image is loaded into the main controller's random access memory for execution. It can be understood that the firmware image is signed and verified to ensure that it has not been tampered with before loading.

[0099] During firmware runtime, the firmware security module needs access to a key key to perform encryption functions of the encryption processing module. The firmware security module reads the device private key used for asymmetric encryption algorithms, or the seed key used to generate the session key, from the physically isolated region of the secure storage chip. In some embodiments, the secure storage chip employs an encryption chip with physical anti-probe capabilities, and the physically isolated region is a protected storage area within the chip. All encryption and decryption operations are performed within the secure region of the chip, including the generation of the symmetric encryption session key, the symmetric encryption of the base weight digital sequence, and the asymmetric encryption of the session key. It can be understood that the secure region is an isolated processing unit within the encryption chip. The reading operations of the private key and seed key are controlled by the secure hardware logic of the encryption chip, and the plaintext form of the private key and seed key only appears within the secure region of the encryption chip. In a specific implementation, when the encryption processing module needs to encrypt the base weight digital sequence, the firmware security module sends an instruction to the secure storage chip, which triggers the chip to derive a symmetric encryption session key using the seed key; the derivation process is completed within the secure region. The symmetric encryption session key generated by the secure storage chip is used to encrypt the incoming base weight digital sequence within a secure area. The encrypted symmetric encryption ciphertext is output to the main controller, while the symmetric encryption session key itself is encrypted by the secure storage chip using its device private key before being output. The private key and seed key exist in ciphertext or hardware-protected form throughout the lifecycle of the secure storage chip; they never leave the secure storage chip.

[0100] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A data-encrypted, anti-cheating intelligent cloud weighing monitoring electronic scale platform, characterized in that, include: The encryption processing module is used to acquire the raw weight signal data generated by each weighing operation of the electronic scale, and generate an encrypted weight data packet after conversion and processing. The symmetric encryption algorithm uses a dynamically generated session key, and the asymmetric encryption algorithm is used to encrypt the session key. The synchronous upload module is used to synchronously upload the encrypted weight data packet, the unique device identification code of the electronic scale, and the weighing operation log containing timestamps to the cloud monitoring server through a preset encrypted communication link. The decryption and restoration module, on the cloud monitoring server, uses an asymmetric private key paired with the electronic scale to decrypt the session key in the encrypted weight data packet, and uses the decrypted session key to symmetrically decrypt the weight data in the encrypted weight data packet to restore the decrypted basic weight digital sequence. The weight calibration module is used to perform calibration and correction processing on the decrypted basic weight digital sequence in the cloud based on pre-stored benchmark sensor calibration curves and dynamic environment compensation models, generating a calibrated standard weight value, including: Based on the unique device identification code of the electronic scale, the corresponding benchmark sensor calibration curve is retrieved from the calibration database stored in the cloud. The benchmark sensor calibration curve describes the mapping relationship between the sensor's original digital signal and the actual physical weight. Extract the temperature and humidity environmental parameters at the moment the weighing occurred from the weighing operation log; The environmental parameters are input into the dynamic environmental compensation model, which is constructed based on the influence of temperature and humidity on sensor sensitivity and zero-point drift. Based on the output of the dynamic environment compensation model, the sensor characteristic correction coefficient for the current environment is calculated. Substitute the value of the decrypted basic weight digital sequence into the calibration curve of the reference sensor to obtain a preliminary weight value. Then multiply the preliminary weight value by the sensor characteristic correction coefficient to obtain the calibrated standard weight value. The audit storage module is used to associate and bind the calibrated standard weight value with the weighing operation log and the device identification code, and store them in the cloud-based anti-tampering audit database to form an irrefutable weighing transaction record.

2. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 1, characterized in that, The process of generating an encrypted weight data packet after conversion includes: The system acquires the raw weight signal data generated by each weighing operation of the electronic scale, and performs analog-to-digital conversion and noise reduction filtering on the raw weight signal data to generate a basic weight digital sequence. The basic weight digital sequence is encrypted by combining a symmetric encryption algorithm with an asymmetric encryption algorithm to generate an encrypted weight data packet. The symmetric encryption algorithm uses a dynamically generated session key, and the asymmetric encryption algorithm is used to encrypt the session key. The step of performing analog-to-digital conversion and noise reduction filtering on the original weight signal data to generate a basic weight digital sequence includes: The analog voltage signal output by the electronic scale sensor is sampled and quantized by a high-precision analog-to-digital converter to generate an initial weight digital sequence. For the initial weight digital sequence, an adaptive sliding window midpoint filtering algorithm is used to remove abnormal jump points introduced by mechanical vibration or transient electromagnetic interference; The filtered sequence is further processed by multi-scale decomposition based on wavelet transform to separate the high-frequency noise component and the low-frequency trend component of the signal. The high-frequency noise component is subjected to threshold processing to suppress random noise, and the processed high-frequency noise component is reconstructed with the low-frequency trend component to obtain a smoothed weight signal sequence. The mean of a stable region is extracted from the smoothed weight signal sequence. The stable region is determined by a preset sequence variance threshold. The mean is the base weight digital sequence.

3. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 2, characterized in that, The process of combining symmetric and asymmetric encryption algorithms to perform hybrid encryption on the basic weight digital sequence to generate an encrypted weight data packet includes: A random number seed is generated locally on the electronic scale, and the random number seed is enhanced using the physical random source of the hardware security module; Based on the enhanced random number seed, a symmetric encryption session key for this weighing session is generated through a key derivation function; The basic weight digital sequence is encrypted using the Advanced Encryption Standard algorithm and the symmetric encryption session key to generate symmetric encrypted ciphertext. Obtain the public key from the asymmetric key pair pre-set on the cloud monitoring server, and use the public key to perform asymmetric encryption on the symmetric encryption session key to generate the encrypted session key; The symmetric encrypted ciphertext, the encrypted session key, and the message authentication code used for integrity verification are encapsulated and packaged to form the encrypted weight data packet.

4. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 3, characterized in that, It also includes a behavior detection module, used to detect weighing cheating behavior through multi-dimensional data analysis, as detailed below: Extract the calibrated standard weight value sequence of the same electronic scale or the same batch of goods from the anti-tampering audit database in chronological order; Statistical analysis was performed on the calibrated standard weight value sequence to calculate the mean and variance of the weight values, and a weight trend graph over time was constructed. The calibrated standard weight value of the current weighing is compared with the average value of the historical sequence. If the deviation exceeds the preset reasonable fluctuation threshold, a weight abnormality event is triggered. Simultaneously, the weighing operation log was analyzed to identify suspicious operation patterns during the weighing process, including abnormally short time intervals between the placement and removal of the weighing tray, and repeated weighing operations before the weighing stabilized. Logically correlate the weight anomaly event with the suspicious operation mode. If they occur simultaneously, generate a cheating suspicion alarm and store the cheating suspicion alarm, related weighing records, and related evidence chain into a special monitoring database.

5. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 4, characterized in that, The identification of suspicious operating patterns during the weighing process includes: The state change sequence of each weighing event is parsed from the weighing operation log, and the state change sequence includes the transition between zero point state, stable state and overload state. Calculate the first time interval from when the weight touches the weighing pan until the weight reading reaches the steady state; Calculate the second time interval during which the weight reading returns from a steady state to zero after the object leaves the weighing pan; If the first time interval or the second time interval is less than the limit time threshold for device physical response and signal processing, it is determined to be an abnormality in rapid pallet placement and removal; The system counts the number of times the weight reading switches between the stable state and the zero point state within a single settlement cycle. If the number of switching exceeds a preset frequent operation threshold, it is determined to be an abnormal repeated weighing operation.

6. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 5, characterized in that, Also includes: The evidence preservation enhancement module is used to enhance the immutability of weighing records based on blockchain technology. The calibrated standard weight value, the weighing operation log, and the device identification code, which have been linked and bound together, are used to generate a structured weighing transaction record. Calculate the hash value of the structured weighing transaction record; The hash value is packaged together with the timestamp and the hash value of the previous block to form the data of a new block; The new block is digitally signed using the private key of the cloud monitoring server; The digitally signed new block is broadcast to a pre-built permissioned blockchain network for consensus verification. Once verified, the new blockchain is connected to a regulatory blockchain composed of data from multiple electronic scales.

7. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 6, characterized in that, The step of packaging the hash value together with the timestamp and the hash value of the preceding block to form a new block includes: Obtain the block header hash value of the latest block on the blockchain, and use it as the hash value of the preceding block; Obtain the current accurate time from a trusted time source within the system, and use it as the authoritative timestamp for this weighing transaction; Create a new data structure that includes at least a data body field, a timestamp field, and a previous hash field; The hash value of the structured weighing transaction record is filled into the data body field, the authoritative timestamp is filled into the timestamp field, and the hash value of the preceding block is filled into the preceding hash field. The new data structure is serialized and encoded to generate the original data packet of the new block.

8. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 7, characterized in that, Also includes: The firmware security module is used to implement secure firmware startup and key protection on the electronic scale. After the electronic scale is powered on, its main controller reads the boot program code and the preset root certificate public key hash value from the read-only memory; Read the signed firmware image and corresponding digital signature from the secure storage chip; The validity of the digital signature is verified using the hash value of the root certificate public key to ensure that the firmware image has not been tampered with; After successful verification, the firmware image is loaded and run; During firmware operation, the device private key used for the asymmetric encryption algorithm, or the seed key used to generate the session key, is read from the physically isolated area of ​​the secure storage chip. All encryption and decryption operations are performed in a secure area within the chip, and the private key and seed key never leave the secure storage chip.

9. The data encryption anti-cheating intelligent cloud weighing supervision electronic scale platform according to claim 8, characterized in that, Also includes: The intelligent analysis module is used for self-evolutionary identification of cheating behavior patterns based on machine learning. Continuously collect all confirmed cheating case data from the special monitoring database to form a cheating sample training set, and at the same time collect normal weighing data to form a normal sample training set. Multidimensional feature vectors are extracted from the cheating sample training set and the normal sample training set. The multidimensional feature vectors include weight deviation features, operation timing features, environmental parameter features and device status features. An ensemble learning classifier is trained using the multidimensional feature vectors, the ensemble learning classifier being a combination of decision trees, support vector machines and neural network base classifiers; The trained ensemble learning classifier is deployed to a cloud-based data analytics engine for real-time analysis of newly uploaded weighing transaction records. Based on the output of the ensemble learning classifier, a cheating risk probability score is calculated for newly occurring weighing behavior, and records with cheating risk probability scores higher than the dynamically adjusted threshold are automatically marked as high-risk records and pushed to the supervisor's interface for review.

Images