CN-122027110-A - Multi-source side information fusion deterministic leakage detection method for automobile password chip

CN122027110ACN 122027110 ACN122027110 ACN 122027110ACN-122027110-A

Abstract

The application discloses a multisource side information fusion deterministic leakage detection method for an automobile password chip, which relates to the field of diagnosis and prediction and comprises the steps of A1, synchronously collecting multisource side information, namely synchronously collecting physical side information signals generated by the automobile password chip when a preset password algorithm is executed by adopting high sampling rate collecting equipment, wherein the physical side information signals comprise power consumption signals and electromagnetic radiation signals, N groups of power consumption traces and electromagnetic radiation traces which are in one-to-one correspondence are obtained, each group of traces corresponds to a random input plaintext, A2, performing time-frequency domain unified conversion, and specifically executing short time Fourier transform STFT on each power consumption trace and each electromagnetic radiation trace obtained in the step A1, wherein the STFT adopts a rectangular window as a window function, the window length is determined in advance according to the frequency characteristic of the physical side information signals, and the time-frequency trace is converted into a time-frequency distribution matrix of a complex frequency domain to obtain a power consumption time-frequency matrix and an electromagnetic time-frequency matrix.

Inventors

FENG QI
GAO DEZHI
GU QIAN
SUN NINGNING
ZHAN YUE
WANG CONG

Assignees

一汽解放汽车有限公司

Dates

Publication Date: 20260512
Application Date: 20260107

Claims (10)

1. The multi-source side information fusion deterministic leakage detection method for the automobile password chip is characterized by comprising the following steps of: step A1, synchronously collecting information of multiple source sides, which specifically comprises the following steps: adopting high sampling rate acquisition equipment to synchronously acquire physical side information signals generated when a vehicle password chip executes a preset password algorithm; the physical side information signals include power consumption signals and electromagnetic radiation signals; Obtaining N groups of power consumption traces and electromagnetic radiation traces which are in one-to-one correspondence, wherein each group of traces corresponds to a randomly input plaintext; step A2, time-frequency domain unified conversion, specifically comprising: Independently executing short-time Fourier transform (STFT) on each power consumption trace and each electromagnetic radiation trace obtained in the step A1; the STFT adopts a rectangular window as a window function, and the window length is determined in advance according to the frequency characteristic of the physical side information signal; converting the time domain trace into a time-frequency distribution matrix of a complex frequency domain to obtain a power consumption time-frequency matrix And electromagnetic time-frequency matrix ; Step A3, multi-source information fusion processing, specifically comprising: Processing the time-frequency distribution matrix by adopting a data level fusion or feature level fusion mode to obtain a fusion result; step A4, multivariate statistical certainty detection specifically comprises the following steps: based on the fusion result of the step A3, detecting whether available leakage related to the secret key exists or not by adopting a corresponding multivariate statistical test method; if the detection result meets the preset judgment condition, judging that the automobile password chip has deterministic leakage.
2. The method for detecting leakage of information fusion certainty on a multi-source side of an automobile cipher chip according to claim 1, wherein the data-level fusion comprises: Time-frequency matrix corresponding to same group of trace And electromagnetic time-frequency matrix Performing element-level complex multiplication with a calculation formula of Generating a fused time-frequency matrix ; Then pair-merging time-frequency matrix Performing inverse short-time Fourier transform ISTFT to obtain a fusion enhanced time domain trace; the fusion enhanced time domain trace serves as a data level fusion result.
3. The method for detecting leakage of information fusion certainty on a multi-source side of an automobile password chip according to claim 2, further comprising: corresponding to data level fusion, the multivariate statistical certainty detection in the step A4 adopts a strategy of F-test statistical test, and specifically comprises the following steps: Step A41a, generating 256 key guesses for the output bytes of the preset cryptographic algorithm S-box, and calculating the intermediate value corresponding to each fusion enhanced time domain trace ; Step A41b, according to the intermediate value N fusion enhanced time domain traces are divided into 9 groups corresponding to hw=0, 1,..8; calculating the ratio of the inter-group variance to the intra-group variance of each sampling point to obtain F statistics; step A41c, calculating a p-value corresponding to the F statistic; If there is at least one of And if the p-value of at least one sampling point is smaller than the preset significance level, judging that deterministic leakage exists.
4. The method for detecting leakage of information fusion certainty of multiple source sides facing to automobile cipher chip according to claim 1, wherein the feature level fusion comprises: for each time window t, from a time-frequency-consuming matrix And electromagnetic time-frequency matrix Respectively screening high signal-to-noise ratio frequency components with the amplitude larger than a preset signal-to-noise ratio threshold value; Combining all the screened frequency components to form a multidimensional feature vector; the multidimensional feature vector is used as a feature level fusion result.
5. The method for detecting leakage of information fusion certainty on a multi-source side of an automobile password chip according to claim 4, further comprising: and (3) carrying out feature level fusion, wherein the multivariate statistical certainty detection in the step (A4) adopts a strategy of Wilk' sLambda-test statistical test, and the method specifically comprises the following steps: Step A42a, generating 256 key guesses for the preset cipher algorithm S-box output bytes Calculating the intermediate value corresponding to each multidimensional feature vector ; Step A42b, according to the intermediate value The N multidimensional feature vectors are divided into 9 groups, corresponding to hw=0, 1,..8; calculating the determinant ratio of the scattering matrix in the group to the total scattering matrix to obtain Wilk' sLambda statistic and converting the statistic into F distribution values; Step A42c, calculating p-value corresponding to the converted F distribution value, if at least one exists If the p-value is smaller than the preset significance level, determining that deterministic leakage exists, and utilizing the statistical correlation of each frequency component in the multidimensional feature vector.
6. The automobile-password-chip-oriented multisource-side information fusion deterministic leak detection method according to any one of claims 1 to 5, further comprising: the preset cipher algorithm comprises an AES algorithm or an SM4 algorithm; performing element-level complex multiplication includes hadamard product.
7. The multi-source side information data level fusion deterministic leakage detection method for the automobile password chip is characterized by comprising the following steps of: step B1, synchronously collecting multi-source signals, which specifically comprises the following steps: adopting high sampling rate acquisition equipment to synchronously acquire physical side information signals generated by an automobile password chip in the process of executing a preset password algorithm; the physical side information signals include power consumption signals and electromagnetic radiation signals; Acquiring N groups of power consumption traces and electromagnetic radiation traces which are in one-to-one correspondence, wherein each group of traces corresponds to a randomly input plaintext; Step B2, time-frequency domain conversion, specifically comprising: independently executing short-time Fourier transform (STFT) on each power consumption trace and each electromagnetic radiation trace obtained in the step B1; the STFT adopts a rectangular window as a window function, and the window length is determined in advance according to the frequency characteristic of the physical side information signal; converting the trace of the time domain into a time-frequency distribution matrix of the complex frequency domain to obtain a power consumption time-frequency matrix And electromagnetic time-frequency matrix ; Step B3, frequency domain data fusion, specifically comprising: Based on the power consumption time-frequency matrix corresponding to the same group of traces obtained in the step B2 And electromagnetic time-frequency matrix Performing element-level complex multiplication with a calculation formula of ; Enhancing in-phase leakage signals by this operation, canceling random phase noise, generating a fused time-frequency matrix ; And B4, reconstructing a time domain signal, wherein the method specifically comprises the following steps: for the one obtained in step B3 Performing inverse short-time Fourier transform ISTFT, and converting the fusion information of the complex frequency domain back to the time domain to obtain a fusion enhanced time domain trace with the same length as the original trace; Repeating the steps B3 to B4 for all N groups of traces to obtain N fusion enhanced time domain traces; Step B5, F-test deterministic leak detection, specifically comprising: Step B5a, determining the key hypothesis and the intermediate value, specifically including: selecting an intermediate value depending on a partial key in the preset cipher algorithm, wherein the intermediate value comprises a first round S-box output byte of an AES algorithm or a corresponding round S-box output byte of an SM4 algorithm; Guessing the 8-bit key corresponding to the intermediate value to generate 256 key guesses ; Step B5B, grouping traces, specifically including: For each of Calculating intermediate values respectively corresponding to N fusion enhanced time domain traces ; According to intermediate values The N fusion enhanced time domain traces are divided into 9 groups, and the value of the hamming weight corresponding to the grouping is hw=0, 1, 8; and B5c, F statistic calculation, specifically comprising the following steps: Calculating the ratio of the inter-group variance to the intra-group variance of 9 groups for each sampling point of each fusion enhanced time domain trace to obtain F statistics corresponding to the sampling point; And step B5d, leakage judgment, specifically comprising: calculating a corresponding p-value according to the F statistic; If there is at least one of And if the p-value of at least one sampling point is smaller than the preset significance level, judging that the available deterministic leakage related to the intermediate value exists in the automobile password chip.
8. The multi-source side information feature level fusion deterministic leakage detection method for the automobile password chip is characterized by comprising the following steps of: Step C1, synchronous acquisition and time-frequency conversion of multi-source signals, which specifically comprises the following steps: Step C1a, synchronously acquiring power consumption signals and electromagnetic radiation signals generated by an automobile password chip in the process of executing a preset password algorithm by adopting high sampling rate acquisition equipment to obtain N groups of power consumption traces and electromagnetic radiation traces which are in one-to-one correspondence, wherein each group of traces corresponds to a randomly input plaintext; Step C1b, independently performing a short-time fourier transform STFT for each power trace and each electromagnetic radiation trace; the STFT uses a rectangular window as a window function, the window length is predetermined according to the signal frequency characteristic, and the power consumption time-frequency matrix of the complex frequency domain is obtained through conversion And electromagnetic time-frequency matrix ; Step C2, constructing a multidimensional feature vector, which specifically comprises the following steps: for each time window t, a time-frequency matrix is time-frequency-matrix from power consumption And electromagnetic time-frequency matrix Respectively screening frequency components with higher signal-to-noise ratio; the screening standard is that the amplitude of the frequency component is larger than a preset signal-to-noise ratio threshold; combining all the screened frequency components to form a multidimensional feature vector; repeating the operation for each time window t of all N groups of traces to obtain N corresponding multidimensional feature vectors; step C3, wilk' sLambda-test multidimensional deterministic detection, specifically comprising: Step C3a, key hypothesis and vector grouping, specifically includes: Selecting intermediate values of the dependent partial keys in the preset cipher algorithm, including the first round S-box output bytes of the AES algorithm or the corresponding round S-box output bytes of the SM4 algorithm, to generate 256 key guesses ; For each of Calculating intermediate values respectively corresponding to the N multidimensional feature vectors According to intermediate values The hamming weight HW of (1) divides the N multidimensional feature vectors into 9 groups, and the value of the hamming weight corresponding to the grouping is hw=0, 1..8; Step C3b, wilk' sLambda-test statistic calculation, specifically comprises: For each time window t, calculating an intra-group spreading matrix and a total spreading matrix of 9 groups of feature vectors, obtaining Wilk' sLambda-test statistic by the ratio of determinant of the intra-group spreading matrix and the total spreading matrix, and converting the statistic into a statistic value under F distribution; step C3C, leak determination, specifically including: Calculating corresponding p-value according to the converted F distribution statistical value, if at least one exists And if the p-value is smaller than the preset significance level, judging that the available deterministic leakage exists in the automobile password chip, and utilizing the statistical correlation of each frequency component in the multidimensional feature vector in the detection process.
9. The multi-source side information fusion deterministic leak detection system facing the automobile password chip is characterized by comprising: The multisource side information synchronous acquisition module is used for synchronously acquiring physical side information signals generated when the automobile password chip executes a preset password algorithm by adopting high sampling rate acquisition equipment; the physical side information signals include power consumption signals and electromagnetic radiation signals; Obtaining N groups of power consumption traces and electromagnetic radiation traces which are in one-to-one correspondence, wherein each group of traces corresponds to a randomly input plaintext; the time-frequency domain unified conversion module is used for independently executing short-time Fourier transform (STFT) on each power consumption trace and each electromagnetic radiation trace; the STFT adopts a rectangular window as a window function, and the window length is determined in advance according to the frequency characteristic of the physical side information signal; converting the time domain trace into a time-frequency distribution matrix of a complex frequency domain to obtain a power consumption time-frequency matrix And electromagnetic time-frequency matrix ; The multisource information fusion processing module is used for processing the time-frequency distribution matrix in a data level fusion or feature level fusion mode to obtain a fusion result; the multi-element statistical deterministic detection module is used for detecting whether available leakage related to the secret key exists or not by adopting a corresponding multi-element statistical detection method; if the detection result meets the preset judgment condition, judging that the automobile password chip has deterministic leakage.
10. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus; A computer program is stored in the memory, which when executed by the processor causes the processor to perform the steps of the multi-source side information fusion deterministic leak detection method for an automotive cryptographic chip according to any one of claims 1 to 6.

Description

Multi-source side information fusion deterministic leakage detection method for automobile password chip Technical Field The application relates to the field of communication encryption, in particular to a multi-source side information fusion deterministic leakage detection method for an automobile password chip, a multi-source side information data level fusion deterministic leakage detection method for an automobile password chip, a multi-source side information feature level fusion deterministic leakage detection method for an automobile password chip, a multi-source side information fusion deterministic leakage detection system for an automobile password chip, electronic equipment, a storage medium and a diagnosis prediction platform. Background The global automotive industry is undergoing a profound technological transformation, the core driving force of which is the tendency of "quadrupling", i.e. motorization, networking, intellectualization and sharing. In this trend, modern automobiles are no longer stand alone mechanical devices, but rather highly integrated mobile smart terminals. The vehicle performs real-time information interaction with external environments (such as other vehicles, infrastructure and cloud servers) through a vehicle-mounted self-organizing network and a V2X communication technology. The high networking and intellectualization improves the driving experience and the traffic efficiency, and simultaneously leads the attack surface of the automobile to be greatly enlarged. The communication links of the vehicle, the on-board systems, the mobile application interfaces, etc. may be potential portals for hacking. Therefore, constructing a vehicle information security system with deep defense, safety and reliability has become a key premise and a core challenge for the development of the automobile industry. In addressing the security challenges described above, cryptographic techniques play a role as a basis for information security. The method is a core technical means for guaranteeing confidentiality, integrity, availability and non-repudiation of data in the vehicle-mounted network. In particular, cryptographic technology is used ubiquitously in automotive electronic systems. In a vehicle-mounted network (such as a CAN bus or a vehicle-mounted ethernet), communication messages between ECUs (electronic control units) need to be verified through a cryptographic mechanism such as a Message Authentication Code (MAC) and the like so as to prevent malicious instructions from being injected and data from being forged. Before the OTA software update package is issued and installed, the OTA software update package must be subjected to digital signature verification to ensure that the OTA software update package is credible in source and not tampered with, and malicious firmware is prevented from being implanted. For security diagnostics and access control, access to the vehicle diagnostic interface (OBD) requires authentication based on a cryptographic request-response mechanism to prevent unauthorized access and malicious diagnostic operations. To meet the performance and security requirements of the vehicle-rule class, these core cryptographic functions (e.g., AES, SM4, etc. encryption algorithms) are typically integrated in a dedicated Hardware Security Module (HSM) or high-performance SoC chip. The hardware implementation mode of the ECU chip aims to provide a trusted execution environment and resist the attack of a software layer. Although the cryptographic algorithm is implemented in the hardware chip, the conventional network attack and software cracking can be effectively resisted, a new and more concealed physical attack threat, namely Side channel analysis (Side-CHANNELANALYSIS, SCA), is introduced. Unlike traditional cryptographic methods (e.g., differential analysis, linear analysis) that attempt to mathematically find the theoretical vulnerability of the algorithm itself, side channel analysis completely bypasses the mathematical complexity of the algorithm, whose goal of attack is the "physical implementation" of the cryptographic algorithm on specific hardware. When the encryption or decryption operation is performed on the cipher chip, the internal transistor switch of the cipher chip can generate tiny physical information leakage which changes along with operation data. These leaks exist in a variety of forms, such as energy consumption, electromagnetic radiation, and the like. Therefore, it is a critical and challenging task to perform side-channel security assessment of cryptographic devices, i.e., active detection of whether there is information leakage that can be utilized, at various stages of chip design, manufacturing, and authentication. It is against this background that existing leak detection techniques have evolved. Disclosure of Invention The invention aims to provide a power battery pre-charge fault diagnosis method, a power battery pre-charge fault diagnosi