Training method and device of side channel attack identification model, and prediction method and device

Abstract

This specification provides a method and apparatus for training a side-channel attack identification model, as well as a method and apparatus for predicting side-channel attacks. In this method, a device is deployed as a simulated target device, and a side-channel information acquisition device is deployed within a predetermined distance range outside the device. The side-channel information acquisition device collects side-channel information generated during device operation; it collects first side-channel information generated when the device is running but the acquisition device is not running; it collects second side-channel information generated when both the device and the acquisition device are running. Using the first and second side-channel information, a side-channel attack identification model is trained. This model is then used to predict side-channel attacks. This specification's embodiments can more accurately predict whether a side-channel attack has occurred against a device.

Description

Technical Field

[0001] This specification relates to electronic information technology in one or more embodiments, and more particularly to a training method and apparatus for a side-channel attack identification model, and a method and apparatus for predicting side-channel attacks. Background Technology

[0002] With the continuous development of computer technology, side-channel attacks (SCA) have emerged. A side-channel attack refers to an attack that utilizes side-channel information. Side-channel information refers to physical information generated by a device during operation that can be detected externally, such as sound or temperature. A side-channel attack uses this side-channel information, rather than malicious programs implanted in the device, to crack various information used by the device. For example, by using the electromagnetic information generated externally during encryption by the device's encryption chip, a side-channel attack can be used to crack the key, attacking a device like an ATM and stealing its key. Another example is using the sound information generated by the device's keyboard when a user enters their password, a side-channel attack can be used to decipher the password, leading to privacy breaches.

[0003] Therefore, how to detect side-channel attacks is an urgent problem to be solved. Summary of the Invention

[0004] This specification describes, through one or more embodiments, a training method and apparatus for a side-channel attack identification model, and a prediction method and apparatus for side-channel attacks, capable of detecting whether a side-channel attack has occurred against a device.

[0005] Based on the first aspect, a training method for a side-channel attack identification model is provided, including:

[0006] A device is deployed as a simulated target device, and a side-channel information acquisition device is deployed within a preset distance range outside the device; wherein, the side-channel information acquisition device is used to collect the side-channel information generated by the device during operation.

[0007] The first side channel information generated when the device is running and the side channel information acquisition device is not running is collected;

[0008] Collect second side channel information generated when the equipment is running and the side channel information acquisition device is also running;

[0009] A side-channel attack identification model is trained using the first and second side-channel information.

[0010] The step of arranging the side channel information acquisition device within a preset distance range outside the device includes: arranging the side channel information acquisition device at at least two locations within the preset distance range outside the device.

[0011] The acquisition of the second side channel information generated during device operation and the side channel information acquisition device also operation includes:

[0012] The second side channel information is collected at least twice, when the device is running and the side channel information acquisition device is running at at least two locations.

[0013] The provision of side channel information acquisition devices within a preset distance range outside the device includes: arranging at least two types of side channel information acquisition devices within a preset distance range outside the device.

[0014] The acquisition of the second side channel information generated during device operation and the side channel information acquisition device also operation includes:

[0015] For each type of side channel information acquisition device, when the device is running, that type of side channel information acquisition device is activated, while other side channel information acquisition devices are deactivated, and the second side channel information generated at this time is acquired.

[0016] The side channel information acquisition device includes a sound sensor, and the side channel information includes sound information.

[0017] And / or,

[0018] The side-channel information acquisition device includes a temperature sensor, and the side-channel information includes temperature information.

[0019] And / or,

[0020] The side channel information acquisition device includes a power consumption acquisition device, and the side channel information includes power consumption information.

[0021] And / or,

[0022] The side channel information acquisition device includes an electromagnetic signal acquisition device, and the side channel information includes electromagnetic information.

[0023] And / or,

[0024] The side-channel information acquisition device includes an optical signal acquisition device, and the side-channel information includes color information.

[0025] Each acquisition of first side channel information includes: acquiring multiple consecutive first side channel information generated during device operation and when the side channel information acquisition device is not in operation within more than one consecutive sampling period;

[0026] Each acquisition of second side channel information includes: acquiring multiple consecutive second side channel information generated during the operation of the device and the operation of the side channel information acquisition device within more than one consecutive sampling period;

[0027] After collecting side-channel information and before training the side-channel attack identification model, the process further includes:

[0028] For each acquisition of first-side channel information, the difference between every two adjacent first-side channel information is calculated; the calculated, time-series continuous differences are used to form a first time-series feature sequence.

[0029] For each acquisition of second-side channel information, the difference between every two adjacent second-side channel information is calculated; the calculated, time-series continuous differences are used to form a second time-series feature sequence.

[0030] The step of training a side-channel attack identification model using first side-channel information and second side-channel information includes:

[0031] A side-channel attack identification model is trained using the first and second time-series feature sequences.

[0032] Based on the second aspect, a method for predicting side-channel attacks is provided, including:

[0033] When the device is running, it obtains side channel information within a preset distance range.

[0034] Using the obtained side-channel information and the side-channel attack identification model, it is predicted whether there is a side-channel attack on the device.

[0035] Wherein, obtaining the side channel information within the preset distance range of the device includes: obtaining multiple consecutive side channel information within the preset distance range of the device collected in more than one consecutive sampling period;

[0036] The method of using the obtained side-channel information and the side-channel attack identification model to predict whether a side-channel attack exists on the device includes:

[0037] Calculate the difference between each pair of adjacent side channel information;

[0038] A temporal feature sequence is formed by using the calculated, sequentially continuous differences.

[0039] The time-series feature sequence is input into the side-channel attack identification model to obtain the prediction result output by the side-channel attack identification model. The prediction result indicates whether a side-channel attack on the device has occurred.

[0040] According to the third aspect, a training device for a side-channel attack identification model is provided, wherein the device includes:

[0041] The side channel information collection module is configured to collect first side channel information generated when the device is running and the side channel information collection device is not running; and to collect second side channel information generated when the device is running and the side channel information collection device is also running.

[0042] The training execution module is configured to train a side-channel attack identification model using the first side-channel information and the second side-channel information.

[0043] According to the fourth aspect, a side-channel attack prediction device is provided, wherein the device includes:

[0044] The side channel information acquisition module obtains side channel information within a preset distance range of the device when the device is running;

[0045] The prediction execution module uses the obtained side-channel information and the side-channel attack identification model to predict whether there is a side-channel attack on the device.

[0046] According to a fifth aspect, a computing device is provided, including a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method described in any embodiment of this specification.

[0047] Because attackers need to use a side-channel information acquisition device when launching a side-channel attack, and this device is also a device that generates side-channel information. In other words, when a side-channel attack occurs, both the side-channel information acquisition device and the device generate side-channel information; both generate side-channel information. However, if no side-channel attack occurs, there is no side-channel information generated by the acquisition device; only the device generates it. The method and apparatus provided in this specification utilize this characteristic. They train a model using side-channel information generated only when the device is running, and side-channel information generated when both the device and the acquisition device are running simultaneously. The model learns the difference between these two types of side-channel information, thereby distinguishing the changes in side-channel information caused by the superimposed acquisition device during an attack. This allows the model to accurately predict whether a side-channel attack has occurred against a device. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of the system architecture used in one embodiment of this specification.

[0050] Figure 2 This is a flowchart of a training method for a side-channel attack identification model in one embodiment of this specification.

[0051] Figure 3 This is a flowchart of a side-channel attack prediction method in one embodiment of this specification.

[0052] Figure 4 This is a schematic diagram of the training device for a side-channel attack identification model in one embodiment of this specification.

[0053] Figure 5 This is a schematic diagram of the structure of a side-channel attack prediction device in one embodiment of this specification. Detailed Implementation

[0054] The solution provided in this specification will now be described with reference to the accompanying drawings.

[0055] First, it should be noted that the terminology used in the embodiments of this invention is for the purpose of describing specific embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0056] To facilitate understanding of the methods provided in this manual, the system architecture involved and applicable to this manual is first described. For example... Figure 1 As shown, the system architecture mainly includes: a side-channel information acquisition device, a device under test, and a side-channel attack prediction device.

[0057] The device under test can be any type of device, such as a mobile phone, computer, server, router, Internet of Things (IoT) device, private cloud device, embedded device, or some industrial control equipment of nuclear power plants.

[0058] A side-channel information acquisition device is used to collect side-channel information of the device under test (DUT) and send the collected side-channel information to a side-channel attack prediction device. The side-channel information acquisition device varies depending on the type of side-channel information being collected. For example, if the side-channel information to be collected includes sound, the acquisition device may include a sound sensor such as a capacitive acoustic sensor. Conversely, if the side-channel information to be collected includes electromagnetic signals, the acquisition device may include an antenna.

[0059] Side-channel attack prediction devices can be located on servers in the cloud.

[0060] It should be understood that Figure 1 The number of side-channel information acquisition devices, devices under test, and side-channel attack prediction devices shown in the diagram is merely illustrative. Any number can be selected and deployed as needed.

[0061] In the embodiments of this specification, a side-channel attack identification model is first trained, and then the side-channel attack identification model can be used to predict whether a side-channel attack has occurred when a device is running.

[0062] The training method for the side-channel attack identification model and the prediction method for side-channel attacks in the embodiments of this specification are described below.

[0063] Figure 2 This is a flowchart illustrating a training method for a side-channel attack identification model in one embodiment of this specification. The main entity executing this method is the training device for the side-channel attack identification model. It is understood that this method can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities. See also... Figure 2 The method includes:

[0064] Step 201: Deploy a device as a simulated attacked device, and deploy a side channel information acquisition device within a preset distance range outside the device; wherein, the side channel information acquisition device is used to collect the side channel information generated by the device during operation.

[0065] Arrangement 203: Collect the first side channel information generated when the equipment is running and the side channel information collection device is not running.

[0066] Step 205: Collect the second side channel information generated when the device is running and the side channel information acquisition device is also running.

[0067] Step 207: Use the first side channel information and its label, the second side channel information and its label to train a side channel attack identification model.

[0068] Analysis of the characteristics of side-channel attacks reveals that to launch a side-channel attack, an attacker needs to use a side-channel information acquisition device to collect side-channel information generated during device operation. By analyzing this acquired side-channel information, the attacker can steal relevant content and carry out the attack. For example, when a computer is running and a user is entering a password, an attacker can use a side-channel information acquisition device—specifically a sound sensor—located one meter away from the computer to collect the side-channel information (sound information) generated when the user types the password. By analyzing this sound information, the attacker can steal the password. Because a side-channel attack requires the use of a side-channel information acquisition device, and this device is also a device, it also generates side-channel information. In other words, when a side-channel attack occurs, both the side-channel information acquisition device and the device generate side-channel information; both generate side-channel information. However, if no side-channel attack occurs, there is no side-channel information generated by the side-channel information acquisition device; only the device generates side-channel information. Figure 2 The process shown utilizes this characteristic, using side-channel information generated only when the device is running, and side-channel information generated when the device and the side-channel information acquisition device are running simultaneously, to train the model. The model learns the difference between these two types of side-channel information, thereby distinguishing the changes in side-channel information caused by the superimposed side-channel information acquisition device when an attack occurs.

[0069] Each step will be explained separately below.

[0070] First, for step 201: Deploy a device, denoted as device 1, as a simulated attacked device, and deploy a side channel information acquisition device within a preset distance range outside device 1; wherein, the side channel information acquisition device is used to collect the side channel information generated by device 1 during operation.

[0071] It is understandable that step 201 involves setting up a simulated environment for side-channel attacks. Subsequently, the side-channel attack identification model will be trained using the side-channel information collected in this simulated environment.

[0072] In step 201, when arranging the side-channel information acquisition device, the following requirements should be considered:

[0073] 1. The side-channel information acquisition device is positioned within a preset distance range outside device 1 (the simulated attacked device). This preset distance range is set based on empirical values ​​to ensure that the distance between the side-channel information acquisition device and the attacked device during an actual attack falls within this preset distance range. This preset distance range is set based on the typical distance during an attack and the type of side-channel information acquisition device. For example, if the attack is to simulate using sound as a side-channel information source, and the side-channel information acquisition device is a sound sensor, such as a capacitive sensor, then the preset distance range, for example, 3-7 meters, can be determined based on the typical distance between the sound sensor and the attacked device during a side-channel attack using this type of sound sensor, for example, 5 meters. Similarly, if the attack is to simulate using color, i.e., light signals, and the side-channel information acquisition device is a light sensor, such as a capacitive sensor, then the preset distance range, for example, 5-15 meters, can be determined based on the typical distance between the light sensor and the attacked device during a side-channel attack using this type of light sensor, for example, 10 meters.

[0074] 2. When device 1 is running, the side-channel information acquisition device is only used to collect the side-channel information generated by device 1 (i.e., the simulated attacked device) during operation. In order to ensure that the side-channel information generated by the side-channel information acquisition device in the simulated environment is only generated because the side-channel information acquisition device is used to collect the side-channel information generated by device 1, and to avoid interference, the side-channel information acquisition device is only used to collect the side-channel information generated by device 1 during operation and cannot be used to collect information from other devices.

[0075] 3. Optionally, various types of side-channel information acquisition devices may be deployed.

[0076] There are various types of side-channel information, and attacks can utilize any type of side-channel information acquisition device. Therefore, to simulate more side-channel attack scenarios and further improve the training effect of the model, various types of side-channel information acquisition devices can be deployed simultaneously when building the simulation environment for side-channel attacks, such as sound sensors, temperature sensors, power consumption acquisition devices, electromagnetic signal acquisition devices, and optical signal acquisition devices.

[0077] This third requirement is not mandatory; it is only required to improve the training performance of the model.

[0078] 4. Optionally, a side-channel information acquisition device is deployed at at least two locations. Accordingly, a simulated attack is performed at each location, which means that the second side-channel information at the corresponding location needs to be acquired.

[0079] In an actual side-channel attack, the distance between the actual side-channel information acquisition device A used in the attack and the attacked device is denoted as distance 1. In the simulation environment described above, the distance between the side-channel information acquisition device B used in the simulated attack and the simulated attacked device 1 is denoted as distance 2. In reality, distance 1 and distance 2 are highly likely to be different. When the first requirement for deploying the side-channel information acquisition device is met, the impact of this difference can be largely eliminated. To further eliminate the impact of this difference, when setting up the simulation environment, the same side-channel information acquisition device can be deployed multiple times at at least two locations within a preset distance range outside device 1. This simulates more attack scenarios and eliminates the impact of different side-channel information strengths caused by distance.

[0080] This fourth requirement is not mandatory; it is only required to improve the training performance of the model.

[0081] Next, for step 203: collect the first side channel information generated when device 1 is running and the side channel information collection device is not running.

[0082] Here, the side channel information is generated when the acquisition device 1 is powered on and running, and the side channel information acquisition devices are not powered on, thus obtaining the side channel information that is not affected by the side channel information acquisition devices.

[0083] In the embodiments of this specification, the first side channel information may be sound information, such as the sound generated by the operation of the fan inside device 1. In this case, step 203 of collecting the first side channel information includes: collecting sound information within a preset distance range A (e.g., a range of 1 meter) outside device 1.

[0084] In the embodiments of this specification, the first side channel information may be temperature information. In this case, step 203, collecting the first side channel information includes: collecting temperature information within a preset distance range B (e.g., a range of 10 centimeters) outside the device 1.

[0085] In the embodiments described in this specification, the first side channel information can be power consumption. In this case, step 203, collecting the first side channel information includes collecting the power consumption information of device 1.

[0086] In the embodiments of this specification, the first side channel information can be electromagnetic information. In this case, step 203, collecting the first side channel information includes: collecting electromagnetic information within a preset distance range C (e.g., a range of 50 centimeters) outside the device 1.

[0087] In the embodiments of this specification, the first side channel information can be color information, that is, light information. In this case, step 203, collecting the first side channel information includes: collecting color information within a preset distance range D (e.g., a range of 30 centimeters) outside the device 1.

[0088] In one embodiment of this specification, the characteristics of side-channel attacks can be demonstrated by utilizing the changing features of side-channel information. Therefore, one implementation of step 203 includes: during each acquisition of first side-channel information, acquiring multiple consecutive pieces of first side-channel information generated while device 1 is running and the side-channel information acquisition device is not running within more than one consecutive sampling period. For example, in this acquisition, 100 pieces (referred to as one acquisition) of first side-channel information generated while device 1 is running and the side-channel information acquisition device is not running are acquired consecutively.

[0089] After performing step 203, the change in the first side channel information is further obtained. That is, for each acquired first side channel information, the difference between every two adjacent first side channel information is calculated; the calculated, time-series-continuous differences are used to form a first temporal feature sequence. This first temporal feature sequence reflects the temporal change in the side channel information generated when device 1 is running and the side channel information acquisition device is not running.

[0090] Next, for step 205: collect the second side channel information generated when device 1 is running and the side channel information collection device is also running.

[0091] Here, the side channel information generated when the acquisition device 1 is powered on and running, and the side channel information acquisition device is also powered on and running, is obtained, thus obtaining the side channel information affected by the side channel information acquisition device.

[0092] In the embodiments of this specification, the second side channel information can be audio information. In this case, step 205 of collecting the second side channel information includes: collecting audio information within a preset distance range E outside the device 1.

[0093] In the embodiments of this specification, the second side channel information can be temperature information. In this case, step 205, collecting the second side channel information includes: collecting temperature information within a preset distance range F outside the device 1.

[0094] In the embodiments of this specification, the second side channel information can be power consumption. In this case, step 205 of collecting the second side channel information includes: collecting the power consumption information of both the acquisition device 1 and the side channel information acquisition device, and then adding them together.

[0095] In the embodiments of this specification, the second side channel information can be electromagnetic information. In this case, step 205, collecting the second side channel information includes: collecting electromagnetic information within a preset distance range G outside the device 1.

[0096] In the embodiments of this specification, the second side channel information can be color information, that is, light information. In this case, step 205, collecting the second side channel information includes: collecting color information within a preset distance range H outside the device 1.

[0097] Referring to the above explanation of step 201, to further improve the training effect of the model, if the arrangement of the side-channel information acquisition devices meets the above requirement 3, then step 250 includes: sequentially, for each type of side-channel information acquisition device, when device 1 is running, running that type of side-channel information acquisition device and turning off other side-channel information acquisition devices, and acquiring the second side-channel information generated at this time. For example, if the arranged side-channel information acquisition devices include a sound sensor, a light sensor, and a temperature sensor, then the second side-channel information is acquired three times. In the first acquisition, device 1 is running, the sound sensor is running, and the light sensor and temperature sensor are turned off, and then the second side-channel information generated at this time is acquired; in the second acquisition, device 1 is running, the light sensor is running, and the sound sensor and temperature sensor are turned off, and then the second side-channel information generated at this time is acquired; in the third acquisition, device 1 is running, the temperature sensor is running, and the sound sensor and light sensor are turned off, and then the second side-channel information generated at this time is acquired.

[0098] Referring to the above explanation of step 201, to further improve the training effect of the model, if the arrangement of the side-channel information acquisition device meets the above requirement 4, then step 250 includes: acquiring the second side-channel information generated at least twice when the device is running and the side-channel information acquisition device is running at at least two locations. For example, the side-channel information acquisition device, such as an electromagnetic sensor (which can be an antenna), is arranged at a distance of 50 cm from device 1 in the first instance, at a distance of 1 meter from device 1 in the second instance, and at a distance of 1.5 meters from device 1 in the third instance, thus acquiring the second side-channel information three times. In the first acquisition, device 1 is running, and the electromagnetic sensor is running at a distance of 50 cm from device 1, and then the second side-channel information generated at this time is acquired; in the second acquisition, device 1 is running, and the electromagnetic sensor is running at a distance of 1 meter from device 1, and then the second side-channel information generated at this time is acquired; in the third acquisition, device 1 is running, and the electromagnetic sensor is running at a distance of 1.5 meters from device 1, and then the second side-channel information generated at this time is acquired.

[0099] In one embodiment of this specification, the characteristics of side-channel attacks can be manifested by utilizing the changing features of side-channel information. Therefore, one implementation of step 205 includes: during each acquisition of second side-channel information, acquiring multiple consecutive pieces of second side-channel information generated when the device is running and the side-channel information acquisition device is also running within more than one consecutive sampling period. For example, in this acquisition, 100 pieces (referred to as one acquisition) of second side-channel information generated when device 1 is running and a side-channel information acquisition device, such as a sound sensor, is located at position 1 are acquired consecutively. In the next acquisition, 100 pieces (referred to as one acquisition) of second side-channel information generated when device 1 is running and the sound sensor is located at position 2 are acquired consecutively.

[0100] Accordingly, after executing step 205, the change in the second side channel information is further obtained. That is, for each acquired second side channel information, the difference between every two adjacent second side channel information is calculated; and the calculated, time-series-continuous differences are used to form a second temporal feature sequence. This second temporal feature sequence reflects the temporal change in the side channel information generated when device 1 is running and the side channel information acquisition device is also running.

[0101] Next, for step 207: use the first side channel information and its label, the second side channel information and its label to train a side channel attack identification model.

[0102] The first side-channel information tag includes: no side-channel attack occurred.

[0103] The second side-channel information is labeled with: a side-channel attack has occurred.

[0104] One implementation of step 207 includes training a side-channel attack identification model using a first temporal feature sequence and its label, and a second temporal feature sequence and its label. Specifically, the first temporal feature sequence and its label (the label indicating that no side-channel attack has occurred) can be input as a dataset into the side-channel attack identification model to be trained, and the second temporal feature sequence and its label (the label indicating that a side-channel attack has occurred) can be input as another dataset into the side-channel attack identification model to be trained, so that the side-channel attack identification model can learn based on the temporal changes of the side-channel information generated when no side-channel attack has occurred and the temporal changes of the side-channel information generated when a side-channel attack has occurred.

[0105] Of course, in the above Figure 2In the process shown, if multiple types of side channel information are collected for the first or second side channel information, such as sound information, temperature information and color information collected in multiple acquisitions, then a time-series feature sequence will be formed for each type of side channel information. A side channel attack identification model can be trained based on one or more time-series feature sequences.

[0106] Of course, in the above Figure 2 In the process shown, if the second side channel information is collected in multiple acquisitions, and the second side channel information is collected at multiple locations, then the second side channel information collected at each location will form a second time-series feature sequence. A side channel attack identification model can be trained based on one or more second time-series feature sequences.

[0107] In the embodiments of this specification, the side-channel attack identification model may include a multi-class labeling model, such as random forest, support vector machine, XGBoost, or a deep learning model.

[0108] After training the side-channel attack detection model, it can be used to identify whether a side-channel attack has occurred against a device while it is running. See also Figure 3 Methods for predicting side-channel attacks include:

[0109] Step 301: When a device S is running, obtain the side channel information within the preset distance range of device S.

[0110] Step 303: Using the obtained side-channel information and the side-channel attack identification model, predict whether there is a side-channel attack on device S.

[0111] In one embodiment of this specification, the side channel information obtained in step 301 includes: obtaining multiple consecutive side channel information within a preset distance range of device S in more than one consecutive sampling period, such as multiple consecutive sound information, multiple consecutive color information, multiple consecutive temperature values, etc.

[0112] Accordingly, one implementation of step 303 includes:

[0113] Step 3031: Calculate the difference between each pair of adjacent side channel information.

[0114] Step 3033: Use the calculated, sequentially continuous differences to form a temporal feature sequence.

[0115] Step 3035: Input the time-series feature sequence into the side-channel attack identification model to obtain the prediction result output by the side-channel attack identification model.

[0116] For example, for multiple consecutive sound information collected, the difference between each two adjacent sound information, i.e., the sound sample value, is calculated in time sequence. The sound temporal feature sequence containing multiple differences is generated by using each difference. The sound temporal feature sequence is input into the side-channel attack identification model. The side-channel attack identification model identifies the attack based on the characteristics of the sound temporal feature sequence and outputs a prediction result, such as whether a side-channel attack against device S has occurred or not.

[0117] For example, for a series of electromagnetic information collected, the difference between each two adjacent electromagnetic information, i.e., the electromagnetic sampling value, is calculated in time sequence. The electromagnetic time sequence feature sequence, which includes multiple differences, is generated by using each difference and arranged in time sequence. The electromagnetic time sequence feature sequence is then input into the side channel attack identification model. The side channel attack identification model identifies the attack based on the characteristics of the electromagnetic time sequence feature sequence and outputs the prediction result.

[0118] Of course, in the above Figure 3 In the process shown, if side-channel information is collected multiple times, then multiple time-series feature sequences will be formed. The side-channel attack identification model can identify attacks based on one or more time-series feature sequences.

[0119] In one embodiment of this specification, a training apparatus for a side-channel attack identification model is provided, see [link to relevant documentation]. Figure 4 The device includes:

[0120] The side channel information collection module 401 is configured to collect first side channel information generated when the device is running and the side channel information collection device is not running; and to collect second side channel information generated when the device is running and the side channel information collection device is also running.

[0121] The training execution module 402 is configured to train a side-channel attack identification model using the first side-channel information and the second side-channel information.

[0122] exist Figure 4 In one embodiment of the device shown in this specification, the side channel information acquisition device is respectively arranged at at least two locations within a preset distance range outside the device;

[0123] The side channel information collection module 401 is configured to perform at least two collections of second side channel information generated when the device is running and the side channel information collection device is running at at least two locations.

[0124] exist Figure 4 In one embodiment of the device shown in this specification, at least two types of side-channel information acquisition devices are arranged within a preset distance range outside the device;

[0125] The side channel information collection module 401 is configured to execute;

[0126] For each type of side channel information acquisition device, when the device is running, that type of side channel information acquisition device is activated, while other side channel information acquisition devices are deactivated, and the second side channel information generated at this time is acquired.

[0127] exist Figure 4 In one embodiment of the device shown in this specification,

[0128] The side-channel information acquisition device includes a sound sensor, and the side-channel information includes sound information;

[0129] And / or,

[0130] The side-channel information acquisition device includes a temperature sensor, and the side-channel information includes temperature information.

[0131] And / or,

[0132] The side channel information acquisition device includes a power consumption acquisition device, and the side channel information includes power consumption information.

[0133] And / or,

[0134] The side channel information acquisition device includes an electromagnetic signal acquisition device, and the side channel information includes electromagnetic information.

[0135] And / or,

[0136] The side-channel information acquisition device includes an optical signal acquisition device, and the side-channel information includes color information.

[0137] exist Figure 4 In one embodiment of the apparatus shown in this specification, the side channel information collection module 401 is configured to perform:

[0138] Each time the first side channel information is collected, multiple consecutive first side channel information generated during the operation of the device and the operation of the side channel information collection device are carried out within more than one consecutive sampling period.

[0139] Each time the second side channel information is collected, multiple consecutive second side channel information generated during the operation of the equipment and the operation of the side channel information acquisition device are collected within more than one consecutive sampling period.

[0140] Training execution module 402 is configured to execute:

[0141] For each acquisition of first-side channel information, the difference between every two adjacent first-side channel information is calculated; the calculated, time-series continuous differences are used to form a first time-series feature sequence.

[0142] For each acquisition of second-side channel information, the difference between every two adjacent second-side channel information is calculated; the calculated, time-series continuous differences are used to form a second time-series feature sequence.

[0143] A side-channel attack identification model is trained using the first and second time-series feature sequences.

[0144] In one embodiment of this specification, a side-channel attack prediction device is proposed, see [link to documentation]. Figure 5 This includes:

[0145] The side channel information acquisition module 501 acquires side channel information within a preset distance range when the device is running.

[0146] The prediction execution module 502 uses the obtained side-channel information and the side-channel attack identification model to predict whether there is a side-channel attack on the device.

[0147] exist Figure 5 In one embodiment of the device shown in this specification, the side channel information acquisition module 501 is configured to perform: acquiring multiple consecutive side channel information within a preset distance range of the device, collected in more than one consecutive sampling period;

[0148] Accordingly, the prediction execution module 502 is configured to execute:

[0149] Calculate the difference between each pair of adjacent side channel information;

[0150] A temporal feature sequence is formed by using the calculated, sequentially continuous differences.

[0151] The time-series feature sequence is input into the side-channel attack identification model to obtain the prediction result output by the side-channel attack identification model. The prediction result indicates whether a side-channel attack on the device has occurred.

[0152] This specification provides, in one embodiment, a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods of any embodiment in the specification.

[0153] This specification provides a computing device according to one embodiment, including a memory and a processor, wherein the memory stores executable code, and the processor executes the executable code to perform the method of any embodiment of the specification.

[0154] It is understood that the structures illustrated in the embodiments of this specification do not constitute a specific limitation on the apparatus of the embodiments of this specification. In other embodiments of the specification, the above-described apparatus may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0155] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0156] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using hardware, software, widgets, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium.

[0157] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. Training methods for side-channel attack detection models, including: A device is deployed as a simulated target device, and a side-channel information acquisition device is deployed within a preset distance range outside the device; wherein, the side-channel information acquisition device is used to collect the side-channel information generated by the device during operation. The first side channel information generated when the device is running and the side channel information acquisition device is not running is collected; Collect second side channel information generated when the equipment is running and the side channel information acquisition device is also running; A side-channel attack identification model is trained using the first side-channel information and the second side-channel information. Each acquisition of first side channel information includes: acquiring multiple consecutive first side channel information generated during device operation and when the side channel information acquisition device is not in operation within more than one consecutive sampling period; Each acquisition of second side channel information includes: acquiring multiple consecutive second side channel information generated during the operation of the device and the operation of the side channel information acquisition device within more than one consecutive sampling period; After collecting side-channel information and before training the side-channel attack identification model, the process further includes: For each acquisition of first-side channel information, the difference between every two adjacent first-side channel information is calculated; the calculated, time-series continuous differences are used to form a first time-series feature sequence. For each acquisition of second-side channel information, the difference between every two adjacent second-side channel information is calculated; the calculated, time-series continuous differences are used to form a second time-series feature sequence. The step of training a side-channel attack identification model using first side-channel information and second side-channel information includes: A side-channel attack identification model is trained using the first and second time-series feature sequences.

2. The method according to claim 1, wherein, The step of arranging the side channel information acquisition device within a preset distance range outside the device includes: arranging the side channel information acquisition device at at least two locations within the preset distance range outside the device; The acquisition of the second side channel information generated during device operation and the side channel information acquisition device also operation includes: The second side channel information is collected at least twice, when the device is running and the side channel information acquisition device is running at at least two locations.

3. The method according to claim 1, wherein, The arrangement of the side channel information acquisition device within a preset distance range outside the device includes: arranging at least two types of side channel information acquisition devices within a preset distance range outside the device. The acquisition of the second side channel information generated during device operation and the side channel information acquisition device also operation includes: For each type of side channel information acquisition device, when the device is running, that type of side channel information acquisition device is activated, while other side channel information acquisition devices are deactivated, and the second side channel information generated at this time is acquired.

4. The method according to claim 1, wherein, The side channel information acquisition device includes a sound sensor, and the side channel information includes sound information; And / or, The side-channel information acquisition device includes a temperature sensor, and the side-channel information includes temperature information. And / or, The side channel information acquisition device includes a power consumption acquisition device, and the side channel information includes power consumption information. And / or, The side channel information acquisition device includes an electromagnetic signal acquisition device, and the side channel information includes electromagnetic information. And / or, The side-channel information acquisition device includes an optical signal acquisition device, and the side-channel information includes color information.

5. Methods for predicting side-channel attacks, including: When the device is running, it obtains side channel information within a preset distance range. Using the obtained side-channel information and the side-channel attack identification model, it is predicted whether there is a side-channel attack on the device. The side-channel attack identification model is trained using the training method of any one of claims 1 to 4.

6. The method according to claim 5, wherein, The process of obtaining side channel information within a preset distance range of the device includes: obtaining multiple consecutive side channel information within a preset distance range of the device collected in more than one consecutive sampling period; The method of using the obtained side-channel information and the side-channel attack identification model to predict whether a side-channel attack exists on the device includes: Calculate the difference between each pair of adjacent side channel information; A temporal feature sequence is formed by using the calculated, sequentially continuous differences. The time-series feature sequence is input into the side-channel attack identification model to obtain the prediction result output by the side-channel attack identification model. The prediction result indicates whether a side-channel attack on the device has occurred.

7. A training device for a side-channel attack identification model, wherein, The device includes: The side channel information collection module is configured to collect first side channel information generated when the device is running and the side channel information collection device is not running; and to collect second side channel information generated when the device is running and the side channel information collection device is also running. The training execution module is configured to train a side-channel attack identification model using the first side-channel information and the second side-channel information. The side-channel information collection module is configured to perform the following: Each time the first side channel information is collected, multiple consecutive first side channel information generated during the operation of the device and the operation of the side channel information acquisition device are carried out within more than one consecutive sampling period. Each time the second side channel information is collected, multiple consecutive second side channel information generated during the operation of the equipment and the side channel information acquisition device are also running within more than one consecutive sampling period; The training execution module is configured to execute: For each acquisition of first-side channel information, the difference between every two adjacent first-side channel information is calculated; the calculated, time-series continuous differences are used to form a first time-series feature sequence. For each acquisition of second-side channel information, the difference between every two adjacent second-side channel information is calculated; the calculated, time-series continuous differences are used to form a second time-series feature sequence. A side-channel attack identification model is trained using the first and second time-series feature sequences.

8. A device for predicting side-channel attacks, wherein, The device includes: The side channel information acquisition module obtains side channel information within a preset distance range when the device is running; The prediction execution module uses the obtained side-channel information and the side-channel attack identification model to predict whether there is a side-channel attack on the device. The side-channel attack identification model is trained using the training device for the side-channel attack identification model as described in claim 7.

9. A computing device comprising a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method of any one of claims 1-6.

Images