Artificial intelligence side channel analysis method for block cipher side information pseudo wheel elimination

Abstract

The invention relates to an artificial intelligence side channel analysis method for block cipher side information pseudo wheel elimination, and belongs to the technical field of information security passwords. The method is mainly used for processing a side channel waveform which is generated by adding a pseudo wheel in the encryption process of a block cipher algorithm, interferes real wheel identification and affects a key solution, and aims to accurately identify and extract an effective paragraph of a real encryption wheel in the side channel waveform so as to provide more accurate side channel support for subsequent key recovery and analysis. The main process comprises the steps of firstly reading a waveform trajectory and positioning a key operation position, retaining a peak and suppressing noise by adopting a low-pass filtering method, segmenting a filtered waveform into a plurality of segments, then performing principal component analysis (PCA) dimensionality reduction and unsupervised clustering analysis on all the segmented waveform segments, setting the clustering number to be 2, and finally, determining the waveform trajectory according to the clustering number. Only the interval data corresponding to the determined'real encryption round 'is reserved, and pseudo round elimination is completed. According to the method, plaintext or key information does not need to be predicted, and the efficiency and accuracy of side channel analysis can be effectively improved.

Description

Technical Field

[0001] This invention relates to an artificial intelligence side-channel analysis method for eliminating pseudo-wheels in block cipher side information. Specifically, it involves optimizing the side-channel analysis process of block ciphers by combining signal processing, peak detection, dimensionality reduction and clustering techniques with efficient feature extraction, and belongs to the field of information security and cryptography. Background Technology

[0002] Block cipher algorithms are a widely used and important technique in modern cryptography, primarily used to provide data confidentiality. However, in practical applications, even without directly damaging the encryption device or tampering with the encrypted data, analysts can still exploit physical leaks during the encryption process (such as electromagnetic radiation and power consumption) to deduce sensitive information like the key. This type of analysis based on physical signals is called side-channel analysis, and its emergence poses a serious threat to the security of encryption devices. Therefore, side-channel analysis technology has become an important research topic in the fields of information security and cryptography.

[0003] To address this risk, researchers have proposed various protective measures. One effective measure is the pseudo-wheel technique, which inserts fake wheel calculations before, after, or at random positions during the block cipher process. This prevents analysts from directly locating the real encryption wheel, making side-channel analysis more complex and thus providing a certain level of security for sensitive information. Meanwhile, effectively eliminating pseudo-wheels has become an important research direction in the field of side-channel analysis.

[0004] This invention proposes a pseudo-wheel elimination method based on the waveform difference between pseudo-wheels and real wheels. Summary of the Invention

[0005] This invention proposes an artificial intelligence-based side-channel analysis method for eliminating pseudo-wheels in block cipher side information. This method does not require prior knowledge of plaintext or key information and can effectively improve the efficiency and accuracy of side-channel analysis.

[0006] This invention is achieved using the following technical solutions; An AI-based side-channel analysis method for eliminating pseudo-wheels in block cipher side information includes the following steps: Step 1: Collect the energy waveform leaked by the device running the block cipher algorithm during the encryption process, and locate the area in the waveform that performs operations related to secret information; Step 2: Preprocess the acquired waveform by using a low-pass filter to suppress high-frequency noise and preserve peak characteristics; Step 3: Using the encrypted round feature as the anchor point, perform peak detection on the preprocessed waveform and divide the waveform into several segments; Step 4: Perform principal component analysis (PCA) dimensionality reduction on all segmented waveform fragments; Step 5: Perform unsupervised clustering analysis on the dimensionality-reduced waveform segments. The group with more segments in the clustering results is identified as the "real encrypted wheel" segment, and the other group is identified as the "pseudo wheel" segment. Step 6: Only retain the interval data corresponding to the "real encrypted wheel" to complete the pseudo-wheel elimination.

[0007] The beneficial effects of this invention are as follows: 1. This invention can segment the energy waveform of encryption cryptographic algorithms and effectively distinguish between real wheels and pseudo wheels.

[0008] 2. This method does not require prior knowledge of plaintext or key information, which can effectively improve the efficiency and accuracy of side-channel analysis. Attached Figure Description

[0009] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is the energy waveform achieved by AES with pseudowheel protection in an embodiment of the method of the present invention; Figure 3 This is a schematic diagram of the energy waveform after filtering in an embodiment of the method of the present invention; Figure 4 This is a schematic diagram of effective spikes and segmentation in the embodiments of the method of the present invention; Figure 5 This is a schematic diagram of dimensionality reduction clustering in an embodiment of the method of the present invention; Figure 6 This is a schematic diagram of the pseudo-wheel elimination result in an embodiment of the method of the present invention. Detailed Implementation

[0010] The detailed steps of the method of the present invention will now be described in conjunction with the accompanying drawings and embodiments.

[0011] Example In block cipher algorithms, the Advanced Encryption Standard (AES) is a symmetric encryption algorithm widely used for data encryption and information protection. This embodiment collects the energy waveform of a contact smart card during AES cryptographic operations and constructs a waveform set. The method of this invention is then used to eliminate pseudo-wheels on the waveform set.

[0012] like Figure 1 As shown, an artificial intelligence side-channel analysis method for eliminating pseudo-wheels in block cipher side information includes the following steps: Step 1: Collect the energy waveform leaked by the device running the block cipher algorithm during the encryption process, and locate the area in the waveform that performs operations related to secret information.

[0013] Specifically, this embodiment takes the protection implementation of randomly embedding zero to three pseudo-wheels before and after the AES encryption wheel as an example. The energy waveform of the AES implementation with pseudo-wheel protection is as follows: Figure 2 As shown.

[0014] Step 2: Preprocess the acquired waveform by using a low-pass filter to suppress high-frequency noise and preserve peak features.

[0015] Specifically, the sampling rate and cutoff frequency were set to 1000 and 50, respectively. An FIR low-pass filter was used for filtering, and zero-phase filtering was applied. The filtered result is as follows: Figure 3 .

[0016] Step 3: Using the encrypted round feature as the anchor point, perform peak detection on the preprocessed waveform and segment the waveform based on the peak value.

[0017] Specifically, in the preprocessed waveform, valid negative peaks are selected as markers for encryption rounds by setting amplitude thresholds and minimum peak spacing. The waveform is further divided into multiple independent segments, each segment starting from the position of the valid peak and ending at a preset maximum number of sample points, ensuring that each independent waveform segment has a consistent length and covers as many data points as possible. In this case, the segment length is set to 400.

[0018] The effective peaks and their segmentation diagrams found based on the filtered energy waveform are shown below. Figure 4 As shown.

[0019] Step 4: Perform principal component analysis (PCA) dimensionality reduction on all segmented waveform fragments.

[0020] Specifically, all the obtained waveform segments are extracted and combined into a matrix. The PCA algorithm is used to reduce the dimensionality of the waveform segment data, projecting the original high-dimensional data into a two-dimensional space.

[0021] Step 5: Perform unsupervised clustering analysis on the dimensionality-reduced waveform segments. The group with more segments in the clustering results is identified as the "real encrypted wheel" segment, and the other group is identified as the "pseudo wheel" segment.

[0022] Specifically, for the dimensionality-reduced waveform data, the K-Means clustering algorithm is applied for unsupervised learning. The number of clusters is set to 2, dividing the data into two classes. Each waveform segment is labeled either 0 or 1. Segments labeled 0 represent actual encryption rounds, while segments labeled 1 represent pseudo-rounds. In this embodiment, a scatter plot is used to visualize the segments of different categories. Segments labeled 0 (actual encryption rounds) are represented by dots, while segments labeled 1 (pseudo-rounds) are represented by stars. Figure 5 As shown.

[0023] Step 6: Only retain the interval data corresponding to the "real encrypted wheel" to complete the pseudo-wheel elimination.

[0024] Specifically, an empty array of the same length as the original waveform is created as the modified waveform. Then, all the segmented waveform fragments are iterated over, and for each fragment with a cluster label of 0 (i.e., the real encrypted wheel), its corresponding interval in the new waveform is restored, while other fragments remain at zero. Finally, the waveform after pseudo-wheel elimination is obtained, as shown below. Figure 6 As shown.

Claims

1. An artificial intelligence side channel analysis method for eliminating pseudo-rounds of side information of a block cipher, the key of which lies in distinguishing the waveform characteristics of pseudo-rounds from real rounds, comprising the following steps: Step 1: Collecting the energy waveform leaked by a device running a block cipher algorithm during the encryption process, and locating the area of the waveform where the secret information related operation is performed; Step 2: Preprocessing the collected waveform, using a low-pass filtering method to suppress high-frequency noise and retain the spike characteristics; Step 3: Taking the encryption round characteristics as anchor points, performing peak detection on the preprocessed waveform, and dividing the waveform into several segments; Step 4: Performing principal component analysis (PCA) dimensionality reduction processing on all the segmented waveform segments; Step 5: Performing unsupervised clustering analysis on the dimensionality reduced waveform segments, and determining the class with more segmented segments in the clustering result as "real encryption round" segments, and the other class as "pseudo-round" segments; Step 6: Only retaining the interval data corresponding to the "real encryption round", and completing the pseudo-round elimination.

2. An artificial intelligence side channel analysis method for side information pseudo-round elimination of a block cipher as claimed in claim 1, characterized in that, In step 3, peak detection is performed on the preprocessed waveform; Observe the filtered waveform, there are negative spikes before and after each operation, and the operation time is basically the same, locate the key operation according to the peak value of the spikes before and after the operation and the operation interval; Statistical analysis of the sampling points of the real encryption round operation segments, set the maximum sampling point number as the length of the intercepted segment, and intercept each key operation from the starting point according to this length to form a set of trace segment to be analyzed.

3. The artificial intelligence side-channel analysis method of side information pseudo-round elimination of a block cipher of claim 1, wherein, In step 5, K-Means function is used to perform unsupervised clustering analysis on the dimensionality reduced waveform segments, and real encryption rounds and pseudo-rounds are distinguished according to the clustering results; K-Means algorithm is a commonly used clustering algorithm, which divides data into a specified number of clusters by minimizing the distance between sample points and their corresponding cluster centers; in this step, the K-Means function can automatically divide the dimensionality reduced waveform segments into two categories and label them: the waveform segments of real encryption rounds are labeled as 0, and the pseudo-round waveform segments are labeled as 1.

Images