Password and identity dual-security verification system based on keystroke dynamics
By combining piezoelectric and triboelectric sensing mechanisms in the input device of keystroke dynamics and using neural network models for signal processing, the problems of fragile traditional cryptographic verification systems and complex sensing array structure are solved, and higher identity is achieved Verify accuracy and security.
Patent Information
- Application Number
- CN202411070422.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-15
AI Technical Summary
The existing password-based authentication system is vulnerable to attacks for weak passwords, and the sensing array structure is complex and the wire connection is cumbersome, which makes manufacturing and miniaturization difficult, and the crosstalk problem between sensing elements is difficult to solve.
Using a keystroke dynamics-based input device, combined with piezoelectric and triboelectric sensing mechanisms, using BaTiO3@PVDF piezoelectric film and triboelectric sensing arrays of different materials, the design of flexible interactive array electrodes and PDMS spacer layer to reduce conductors , realize a self-powered keystroke sensing system, and use the 1DNN neural network model for signal processing and user identification.
Improve the accuracy and security of identity verification, cover a wider range of pressure detection, enhance password verification through the dual sensing effect of piezoelectric and triboelectric, reduce hardware resources, and achieve non-invasive dual identity verification.
Smart Images

Figure CN118965315A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of piezoelectric and triboelectric sensing, and in particular to a password and identity dual security verification system based on keystroke dynamics. Background Art
[0002] With the development of communication technology, the Internet has penetrated into every corner of our lives, and network security issues have risen to an important level. It covers all network security related to personal information, health, finance, intellectual property and even national security. Password-based authentication has always been the most popular method for computer security systems to authenticate users. However, password-based authentication systems are vulnerable to weak passwords and are susceptible to many attacks such as dictionary attacks, guessing attacks, and shoulder surfing attacks. When passwords are obtained by third parties, security vulnerabilities are caused. In addition to the use of passwords in network security systems, additional authentication procedures are required to enhance security. In recent years, keystroke dynamics has become a hot topic of research in enhancing user authentication.
[0003] Keystroke feature recognition technology refers to the use of keystroke dynamics to obtain personal behavioral information, such as keystroke force, typing speed, delay and holding time between consecutive keystrokes, which can be used to build a seal for user identification. It has the user's typing behavior pattern, which is unique due to physiological mechanisms, similar to handwriting and signatures, and contains enough information as a biometric identifier. Compared with other behavioral biometric technologies such as voice detection, handwriting, gait movement and gestures, biometric technology based on keystroke dynamics is non-invasive, convenient, cost-effective and capable of continuous identity authentication.
[0004] Although keystroke dynamics depend on external factors such as keyboard size, layout, and mechanism, frequently typed words and strings show more consistency. To enhance network security, dual-factor key dynamics have emerged in authentication. A single sensor device acts as a single security layer in authentication, while a hybrid sensing system can provide more security layers to make up for the defects of a single sensing mode, thereby achieving higher security authentication accuracy. However, most hybrid sensing array structures have complex structures, a large number of sensing elements, electrodes, and interfaces, and limited multi-parameter interaction detection capabilities. And the amount of data transmitted is large. In addition, for the interface of the sensor array, the wiring connection will cause crosstalk. In order to eliminate the crosstalk problem, each sensor element requires a separate electrode and wire connection. Therefore, an n×m sensor array requires n×m wires, which multiplied by the number of wires will bring many troubles to the manufacturing and miniaturization of the sensor array. Summary of the invention
[0005] In order to overcome the above technical problems, the purpose of the present invention is to provide a self-powered numeric keyboard based on the complementary enhancement of triboelectric-piezoelectric sensing mechanism, and then propose a password and identity dual security verification system based on keystroke dynamics to solve the problem that the existing password-based identity authentication system is relatively vulnerable to weak passwords and is easily attacked and cracked by hackers.
[0006] To this end, the first technical solution of the present application discloses an input device based on keystroke dynamics, including a piezoelectric sensor located at the top and a triboelectric sensor array device located at the bottom, wherein:
[0007] The piezoelectric sensor uses BaTiO3@PVDF piezoelectric film as the piezoelectric layer;
[0008] The triboelectric sensor uses different materials as a triboelectric sensing array device, and the bottom electrode of the device uses a flexible interactive array electrode printed by FPCB.
[0009] Furthermore, silver electrodes are deposited on both sides of the piezoelectric layer, and a pure PVDF film is used as a packaging layer. Silver electrodes are deposited on the surface of the bottom packaging layer of the piezoelectric sensor as a positive friction layer. PDMS is used as a spacer layer between the positive friction layer and the triboelectric sensor array device, and the size of the spacer layer is consistent with that of the flexible interactive array electrode.
[0010] Furthermore, the BaTiO3@PVDF piezoelectric film is prepared by the following method:
[0011] S1. Preparation of primary crystallized film: BaTiO3 powder was added to DMF solution and ultrasonically stirred, and then PVDF-HFP powder was added and stirred and degassed to obtain a mixed solution; the mixed solution was cast on an aluminum plate and heated and dried to obtain a formed primary crystallized film;
[0012] S2. Uniaxial stretching: The primary crystallized film is stretched at 58-62°C with a stretching rate of 5;
[0013] S3. Thin film polarization: Use CM2400 epoxy conductive adhesive to stick the aluminum foil on the stretched film and solidify it. Then stick the electrode sheets on both sides of the film for oil bath polarization.
[0014] Furthermore, the oil bath polarization is as follows: the polarization electric field strength is 20 MV / m to 90 MV / m, the polarization strength is increased by 10 MV / m every 8 minutes, and the electric field strength is increased for the next step after pausing for 4 minutes.
[0015] The second technical solution of the present application discloses a password and identity dual verification system based on keystroke dynamics, comprising:
[0016] Keystroke information collection module: including the input device based on keystroke dynamics, used to receive user tapping and input user behavior information and different waveform combinations, realizing the dual sensing functions of password and identity;
[0017] Wireless data acquisition and signal processing module: coupled to the keystroke information acquisition module, used to receive the piezoelectric and frictional impedance output by the keystroke information acquisition module, fuse and add them, output them as digital signals, and transmit them to the PC serial port via wireless;
[0018] Pattern recognition module: It is a 1DNN neural network model, which is used to extract features of password signals from two dimensions: time domain and energy domain, and then fuse them at the feature layer, and finally classify the signals to achieve user identification and password verification.
[0019] Furthermore, the wireless data acquisition and signal processing module includes a fusion signal measurement circuit, which is composed of two small voltage measurement circuits and a three-way adder, an AD sampling module, an STM32 microprocessor and a LoRa wireless transmission module; the small voltage measurement circuit is used to process the source impedance of piezoelectricity and triboelectricity, the three-way adder fuses and adds the piezoelectric and triboelectric signals, and the AD sampling module reads the fused voltage signal as a digital signal; the STM32 microprocessor reads the AD signal at a frequency of 1KHz and transmits the data to the LoRa wireless transmission module, and the LoRa wireless transmission module receives the data at the PC end and transmits it to the PC serial port.
[0020] Furthermore, the 1DCNN neural network model in the pattern recognition module is a four-layer network structure, including two convolutional layers and two fully connected layers. The convolutional layer uses the maximum pooling layer for sampling and the Same method is used for edge filling. The entire model is trained using the random optimization method (Adam) with adaptive momentum.
[0021] And, a password and identity dual verification method based on keystroke dynamics, using the password and identity dual verification system based on keystroke dynamics, comprising:
[0022] The digital signal is processed to extract characteristic information, wherein the digital signal is converted from the piezoelectricity and friction electricity input by the user tapping the input device through the front signal acquisition and signal processing module; and the characteristic information is used to identify and verify the user.
[0023] A computer device, comprising:
[0024] Memory; and
[0025] The processor is configured to execute the password and identity dual verification method based on the instructions stored in the memory.
[0026] A computer-readable storage medium stores computer instructions, which can implement the password and identity dual verification method when executed by a processor.
[0027] Compared with the prior art, the present invention has the following beneficial effects:
[0028] In the password and identity dual security verification system based on keystroke dynamics provided by the present invention, the dual pressure detection system based on piezoelectricity and triboelectricity can cover a wider range of pressure detection. Different triboelectric signals can represent different numeric keys, and passwords with personal unique keystroke dynamic information can be input through different key combinations. The top layer integrates piezoelectric sensing, and the two sensing effects have complementary and enhanced effects, overcoming the limitations of single sensing and improving the accuracy of authentication.
[0029] In the password and identity dual security verification system based on keystroke dynamics provided by the present invention, the self-powered keystroke sensing system combined with the neural network algorithm can accurately verify the password and distinguish users. With self-powered sensing, dual authentication system (password waveform group and keystroke characteristics) and deep learning neural network, this work provides a dual security layer with behavioral biometrics and password verification for network security.
[0030] In the password and identity dual security verification system based on keystroke dynamics provided by the present invention, a longitude and latitude array electrode design method is adopted to successfully reduce the 3×3 sensor array output channels to 1 channel, thereby reducing hardware resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 A schematic diagram of a system for password and identity dual security verification based on keystroke dynamics provided in part of the embodiments of the present invention;
[0032] Figure 2 A schematic diagram of the preparation of the BaTiO3@PVDF piezoelectric film provided in the embodiments of the present invention;
[0033] Figure 3 A piezoelectric simulation model diagram provided in the embodiment of the present invention;
[0034] Figure 4 The triboelectric sensor mechanism analysis and performance characterization diagram provided in the embodiments of the present invention;
[0035] Figure 5 A schematic diagram of a fusion voltage measurement circuit structure of a password and identity dual security verification system based on keystroke dynamics provided in the embodiments of the present invention;
[0036] Figure 6Voltage waveforms of PENG, TENG and PENG-TENG fusion provided in the embodiments of the present invention;
[0037] Figure 7 A schematic diagram of a password and identity dual security verification system based on keystroke dynamics using a neural network model for biometric password verification provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0039] Unless otherwise specifically stated, the relative arrangement, numerical expressions and numerical values of the parts and steps set forth in these embodiments do not limit the scope of the present application. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to those of ordinary skill in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be considered as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so that once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.
[0040] Unless otherwise specified, the meanings of the technical terms in this specification are the same as those generally understood by those skilled in the art. However, if there is any conflict, the definitions in this specification shall prevail.
[0041] The first embodiment of the present application discloses an input device based on keystroke dynamics, including a piezoelectric sensor located at the top and a triboelectric sensor array device located at the bottom, wherein: the piezoelectric sensor uses BaTiO3@PVDF piezoelectric film as the piezoelectric layer; the triboelectric sensor uses different materials as the triboelectric sensor array device, and the bottom electrode of the device uses a flexible interactive array electrode printed by FPCB.
[0042] As a dual security verification system for passwords and identities based on keystroke dynamics, the input device is first required to generate user behavior information and different password waveform combinations during the user's key input process to achieve dual sensing functions of passwords and identities. Therefore, this application first proposes an input device based on keystroke dynamics, taking a keyboard as an example, Figure 1 As shown in Figure 2, it includes a piezoelectric sensor at the top and a triboelectric sensor array device at the bottom. The piezoelectric sensor at the top uses 6cm×6cm BaTiO3@PVDF as the piezoelectric layer. 200nm silver electrodes are deposited on both sides of the piezoelectric film. A commercial 28μm thick pure PVDF film is used as the encapsulation layer. The triboelectric sensor at the bottom uses 9 materials as the friction layer (polydimethylsiloxane (PDMS), silicone (Ecoflex), rubber (Rubber), polyvinylidene fluoride (PVDF), polytetrafluoroethylene (PTFE), polyimide (PI), polycaprolactone (PCL), nylon (PA66), polyurethane (PU)). The friction layer prepared by using different materials can generate different triboelectric signals when tapped. The bottom electrode uses a 3*3 flexible interactive array electrode printed by FPCB process. Only one output interface is needed to collect 9 triboelectric signals, avoiding the crosstalk problem of array sensing. The thickness of the FPCB electrode is 50μm, and the diameter of each electrode is 2cm. Each circular electrode is connected by a 2mm wide rectangular copper. A 0.025μm gold layer is deposited on the electrode surface for anti-oxidation and long-term stability. A 0.5μm silver electrode is immersed on the surface of the bottom packaging layer of the piezoelectric sensor as a positive friction layer. The middle spacer layer uses PDMS with 3×3 circular holes as the spacer layer. The size of the hole is the same as the size of the circular electrode. The size of the spacer layer is 2mm×6cm×6cm. Figure 1 .b and Figure 1 .c.
[0043] The BaTiO3@PVDF piezoelectric film is a composite material prepared by uniformly dispersing barium titanate (BaTiO3) nanoparticles in a polyvinylidene fluoride (PVDF) matrix. This material combines the high piezoelectric coefficient of BaTiO3 with the flexibility and chemical corrosion resistance of PVDF, and has the following characteristics and advantages:
[0044] (1) Good piezoelectric properties. BaTiO3@PVDF piezoelectric film can show good piezoelectric responsiveness in a wide frequency range and is suitable for vibration detection and energy harvesting at various frequencies.
[0045] (2) Mechanical flexibility: the PVDF matrix gives the film good flexibility, allowing the piezoelectric film to be flexibly applied to various curved and complex surfaces;
[0046] (3) Chemical corrosion resistance: PVDF has excellent chemical stability, which enables BaTiO3@PVDF piezoelectric film to maintain stable performance in harsh chemical environments;
[0047] (4) Environmentally friendly. The biodegradable properties of BaTiO3@PVDF piezoelectric film are of great significance for reducing environmental pollution and improving sustainable development.
[0048] (5) Easy to process. PVDF material can be processed into a specific shape, so that the BaTiO3@PVDF piezoelectric film can be completely fitted with the measured surface, improving the sensitivity and accuracy of the sensor;
[0049] In a specific embodiment, the BaTiO3@PVDF piezoelectric film is prepared by the following method: including primary crystallization film preparation, uniaxial stretching, and film polarization;
[0050] The primary crystallization film preparation comprises the following steps:
[0051] Add BaTiO3 powder to 25 ml of DMF solution and stir ultrasonically for 10 minutes. Then add 8 g of PVDF-HFP powder, stir in a planetary mixer for 10 minutes, and degas for 2 minutes;
[0052] The aluminum plate (10cm*20cm*1mm) used for casting the solution was cleaned with alcohol, and then a square area of 16cm×8.5cm was drawn on the surface of the aluminum plate, and the sponge glue was fixed on this area;
[0053] Pour the mixed solution into the area where the sponge glue is located on the aluminum plate, place it on the heating table for baking, preheat the heating table for 15 minutes at 90°C, and remove the sponge glue after 30 minutes. After heating for two hours, the initial crystallized film is obtained;
[0054] The uniaxial stretching comprises the following steps:
[0055] After the primary crystallized film is formed and flattened with an aluminum plate, a region with a thickness of 280 μm to 350 μm, uniform thickness, flat surface and no bubbles or depressions is selected on the film, and a film with a size of 8 cm × 8 cm is cut out;
[0056] Taking the film width midline as the reference, use a marker to draw lines 0.7 cm and 0.5 cm above and below the midline (the line at 0.7 cm is used as the mold clamping line during stretching, and the line at 0.5 cm is used as the reference line). The width of the stretching area is 1.4 cm.
[0057] The film was clamped and fixed on the stretching machine, and then covered with two layers of wooden temperature control box. The inside of the temperature control box was heated with a digital hot air gun to keep the temperature stable at about 60°C (error ±2°C). The stretching stroke was set to 100mm, the stretching rate was set to 10mm / min, the stretching rate was 5, the film thickness before stretching was about 300μm, and the film thickness after stretching was 60μm-90μm;
[0058] The thin film polarization comprises the following steps:
[0059] Cut out a piezoelectric film with a size of 7 cm × 7 cm (along the stretching direction) in the film stretching area, leaving 0.2 cm around for pasting masking tape;
[0060] Use CW2400 epoxy conductive adhesive to stick 15μm thick aluminum foil to the film, flatten it with an aluminum plate and cure it for 24 hours, then remove the masking paper and replace it with insulating tape;
[0061] Stick the two electrodes of the power supply to both sides of the film, fix them with clips, and perform oil bath polarization. The polarization electric field strength is 20MV / m to 90MV / m. Increase the polarization strength by 10MV / m every 8 minutes. Pause for 4 minutes before increasing the electric field strength for the next step.
[0062] It should be noted that the PVDF primary crystallized film obtained by the conventional solution casting method (adding BaTiO3 to the PVDF film by casting) is mainly α-crystalline phase, and needs to be modified to achieve crystal phase transformation and electric dipole orientation. Figure 2 a and Figure 2 As shown in b, after directional stretching and polarization treatment, the structure changes from the isotropic state to the anisotropic state. Comparing the materials before and after adjustment, the PVDF molecular chain configuration tends to change from a non-polar phase (TGTG') to a polar phase (TTTT). The main purpose of directional stretching of the film is to transform the α crystal phase (spherulite structure) inside the film into a piezoelectric β crystal phase (fibrous structure). Because the electric dipole orientation in the initial piezoelectric film is random and irregular, the whole does not exhibit piezoelectricity. The carbon-fluorine dipole moment of the β crystal phase is oriented perpendicular to the molecular chain to form a spontaneous polarization intensity. After further polarization treatment, the dipole moment inside the unit cell is further rotated and oriented to form a fixed carbon-fluorine dipole moment structure perpendicular to the plane of the film. When the film is subjected to external force, the residual polarization intensity changes, and the film exhibits a piezoelectric effect. Adding piezoelectric ceramic molecules to polymer compounds not only maintains the softness of the polymer piezoelectric film, but also has a high piezoelectric coefficient. It is a piezoelectric material with development prospects. As Figure 2 c and Figure 2 As shown in Figure d, TEM shows that the morphology of BaTiO3 nanoparticles is uniformly distributed spheres. Figure 2 e and Figure 2As shown in Figure f, clear distribution of Ba and O elements can be observed. Figure 2 As shown in Figure g, the XRD spectrum shows that the characteristic peaks of the primary crystallized film mainly appear at 18.5°, 20.1° and 26.5°, corresponding to the (100), (110) and (021) crystal planes of the α crystal phase, respectively. The characteristic peaks of the β crystal phase are not obvious, indicating that the crystal phase in the primary crystallized film is mainly the α crystal phase. Figure 2 As shown in h, the XRD spectrum shows that after the initial crystallization film is stretched, there is basically only one characteristic peak at 20.6°, that is, the (110) crystal plane of the β crystal phase, indicating that the α crystal phase is almost completely transformed into the β crystal phase. Figure 2 It can also be seen from the Fourier infrared spectrum shown in Figure 1 that the β-phase has a wave number of 837-841 cm -1 The characteristic absorption peaks in the range of β-phase and β-phase are gradually enhanced after the film is stretched and polarized. The crystal phase transition occurs more in the stretching stage, and the effect of polarization on the crystal phase transition of PVDF film is not obvious.
[0063] In order to further study the piezoelectric performance of the piezoelectric sensor of the present invention, the output performance of the sensor under different pressures was tested. Figure 3 As shown;
[0064] Furthermore, when the piezoelectric sensor is impacted by external force, the output of the piezoelectric sensor is limited by two factors. First, PVDF will be extruded and deformed when impacted by external force. Before deformation, the electric dipoles in the film will be regularly arranged in the film, and the polarization trend is relatively stable. At this time, no electrical signal will be generated. When the film is compressed and deformed, the arrangement of the electric dipoles in the film changes, resulting in a decrease in the polarization strength inside the film, thereby generating an electrical signal;
[0065] Furthermore, the piezoelectric performance of the piezoelectric sensor is also affected by the BaTiO3 nanoparticles. Since the BaTiO3 nanoparticles are wrapped on the surface of the PVDF-HFP film, when the sensor is impacted, there is abundant deformation and strong piezoelectric potential between the BaTiO3 nanoparticles due to mutual squeezing. Under the combined effect of the two effects, the output performance of the piezoelectric sensor is improved;
[0066] Figure 3 a is a simulation model of piezoelectric material being deformed by pressure;
[0067] Figure 3b is the basic piezoelectric performance of the piezoelectric sensor. In order to obtain the piezoelectric characteristics of the sensor, we applied five different forces (0.5N, 1N, 2N, 3N, 4N and 5N) to the sensor at an impact frequency of 2Hz. It can be seen that when the pressure is 0.5N, the average output voltage of the sensor is 0.7V. When the pressure is 5N, the average value of the output voltage is 3V. The data between 0.5-5N also confirms that the voltage of the sensor in the impact mode increases with the increase of external force;
[0068] Figure 3 The voltage waveform changes under different loads of c Figure 3 b The open circuit voltage shows the same trend;
[0069] like Figure 3 As shown in Fig. e, after cycling 5000 times at 1 Hz and 2 N pressure, the voltage amplitude output is stable, indicating that the piezoelectric sensor has high mechanical stability and durability.
[0070] like Figure 4 As shown in the figure, based on the working mechanism of triboelectric sensors, a 3×3 sensor array is constructed. The sensor array consists of 9 types of friction sensors. Figure 4 Analyze 9 irregular triboelectric signal waveform curves. Since triboelectric charging is a dual-material interface phenomenon and strongly depends on the surface contact of the materials, the difference in triboelectric performance phenomena between different materials is due to the different electron cloud states between different materials. Therefore, the triboelectric signal between two given materials is unique. The position in the triboelectric series determines the efficiency of charge exchange. According to the triboelectric series, PTFE has the strongest electronegativity among the 9 materials, followed by PDMS, PI, PVDF, PCL, PU, Ecoflex, Rubber, and PA66 has the strongest electropositivity (electronegativity: PTFE>PDMS>PI>PVDF>PCL>PU>Ecoflex>Rubber>PA66). When testing the triboelectric sensors of each array separately, such as Figure 4 As shown in (a, b, h), the sensing units based on PA66, Rubber and Ecoflex all produced similar “peak-drop” voltage waveforms. Compared with the Ag electrode, PA66, Rubber and Ecoflex have stronger electron loss capabilities; Figure 4The waveforms of (c, d, e, f, g and i) have waveforms in opposite directions (peak inclinations), because PTFE, PDMS, PI, PVDF, PCL and PU have a stronger tendency to obtain electrons than the Ag electrode. Therefore, a "rising peak" voltage waveform is generated. Among the three voltage waveforms with falling peaks, the triboelectric output amplitude of the PA66 (1.25V) sensor is the largest, followed by the Ecoflex (0.86V) and Rubber (0.48V) sensors. Among the six voltage waveforms with rising peaks, the triboelectric output amplitude of the PTFE sensor is the largest (1.65V), followed by PDMS (1.32V), PI (1.19V), PVDF (0.9V), PCL (0.75V), and PU (0.47V). The reason why different sensors output different triboelectric amplitudes is attributed to the differences in the ability of different materials to gain and lose electrons. Each independent sensor generates a different triboelectric signal when in contact / separation, and each voltage signal has a unique waveform and relative amplitude characteristics. In addition, the longitude and latitude array electrode structure eliminates the impact of crosstalk, and each device outputs independently without affecting each other.
[0071] It should be noted that the above is only a specific implementation method of the present application illustrated by way of example, and the scope of protection of the present invention is not limited to this. For example, for the sake of ease of explanation, the triboelectric sensor of this specific implementation method uses 9 materials to collect different friction signals respectively; but it does not mean that only 9 materials can be used. The present application can use a variety of different materials to collect different friction signals according to different input and scenario requirements.
[0072] like Figure 1 A schematic diagram of the input device of the present invention is shown in Figure a, and the keystroke dynamics password authentication system is complementarily integrated by triboelectric and piezoelectric sensors. The user inputs the set password waveform through the keystroke fusion sensor, and the electrical signal is collected by the data acquisition system. The collected user data is then transmitted to the signal processing module for signal processing and then automatically extracted through the neural network algorithm model. Predefined features are used to build a profile database. During the identity authentication / password recognition process, a test profile will be constructed according to the same procedure and cross-referenced with the existing profile database to determine whether the test subject is an authorized user and verify the password.
[0073] As can be seen from the above, the input device based on keystroke dynamics disclosed in the present application solves the problem of high energy consumption of array sensing at the same time. The self-powered hybrid sensor can continuously collect the user's keystroke information in a non-invasive manner. In addition, through structural design, the sensing array composed of multiple independent triboelectric sensors sharing array electrodes overcomes the crosstalk problem of general array sensing. The array sensor can provide a variety of different voltage waveforms to constitute unique information of multiple function keys. When different sensors are pressed, different voltage waves are generated. By setting a password, a signal wave group with a series of different waveform combinations can be generated. The piezoelectric sensing layer on the top provides an enhanced signal for the password waveform, giving the signal multimodality and versatility. The fused waveform has the characteristics of wide bandwidth, multiple signals, and strong robustness. This battery-free keystroke sensing system combined with a neural network algorithm can accurately verify passwords and distinguish users. With self-powered sensing, a dual authentication system (password waveform group and keystroke features) and a deep learning neural network, this work provides a dual security layer with behavioral biometrics and password verification for network security.
[0074] The second embodiment of the present application discloses a password and identity dual verification system based on keystroke dynamics, comprising:
[0075] Keystroke information collection module: including the input device based on keystroke dynamics, used to receive user tapping and input user behavior information and different waveform combinations, realizing the dual sensing functions of password and identity;
[0076] Wireless data acquisition and signal processing module: coupled to the keystroke information acquisition module, used to receive the piezoelectric and frictional impedance output by the keystroke information acquisition module, fuse and add them, output them as digital signals, and transmit them to the PC serial port via wireless;
[0077] Pattern recognition module: It is a 1DNN neural network model, which is used to extract features of password signals from two dimensions: time domain and energy domain, and then fuse them at the feature layer, and finally classify the signals to achieve user identification and password verification.
[0078] In a specific implementation, the wireless data acquisition and signal processing module includes a fusion signal measurement circuit, which is composed of two small voltage measurement circuits and a three-way adder, an AD sampling module, an STM32 microprocessor and a LoRa wireless transmission module; the high-impedance voltage measurement circuit is used to process the source impedance of piezoelectricity and triboelectricity, the three-way adder fuses and adds the piezoelectric and triboelectric signals, and the AD sampling module reads the fused voltage signal as a digital signal; the STM32 microprocessor reads the AD signal at a frequency of 1KHz and transmits the data to the LoRa wireless transmission module, and the LoRa wireless transmission module receives the data at the PC end and transmits it to the PC serial port.
[0079] In a specific implementation, the 1DCNN neural network model in the pattern recognition module is a four-layer network structure, including two convolutional layers and two fully connected layers. The convolutional layer uses a maximum pooling layer for sampling and uses the Same method for edge filling. The entire model is trained using a random optimization method (Adam) with adaptive momentum.
[0080] It should be noted that the fusion signal measurement circuit is composed of two small voltage measurement circuits and an adder circuit. The small voltage measurement circuit is used to collect the voltage of the device, and the three-way adder circuit is used to couple the piezoelectric and triboelectric signals to output. Figure 5 As shown;
[0081] Furthermore, the small voltage measurement circuit is composed of the following two stages: the first stage is an impedance conversion circuit, which can convert a high-impedance voltage signal into a low-impedance output voltage signal; the second stage is a signal coupling circuit, which couples the piezoelectric signal and the friction electric signal into a characteristic signal output.
[0082] The present invention is specifically described below by way of examples.
[0083] The fusion signal output by the fusion signal measurement circuit is tested, such as Figure 6 a-6i shows 9 synchronous waveforms of triboelectric signals and piezoelectric signals, as well as a fusion signal obtained by coupling the two signals;
[0084] The device was placed on a flat surface and tested in an intermittent tapping manner to test three signals (piezoelectric signal, triboelectric signal, fusion signal). When testing one of the devices, the other device was kept open and isolated, and the signals corresponding to the single isolation mechanism, namely PENG, TENG and piezoelectric tribohybrid contact (HYB), were detected to avoid the secondary phenomenon of charge dispersion. In this case, TENG does not represent a complete hybrid effect because the TENG electrodes are completely disconnected, so it is a measure of the triboelectric effect in the piezoelectric tribohybrid contact. Post-processing signal smoothing was performed to reduce interference and noise. The three signals all follow their respective electrode polarities and show different waveforms. Each fusion waveform has a complete piezoelectric waveform and triboelectric waveform, and each waveform feature has been well verified. The 9 fusion signals retain the waveform features of piezoelectric and triboelectric intact;
[0085] It is worth noting that the polarization direction of PVDF and the working mode of TENGs have a significant impact on the electrical output. The trend of synchronous changes in piezoelectric and triboelectric signals helps to enhance the contact potential and achieve better performance.
[0086] Figure 6j(i-iv) describes a schematic diagram of the working principle of the input device of the present application, which includes three stages:
[0087] (1) In the initial state, the finger has not yet touched the input device, the input device is placed flat, and the piezoelectric component has no charge or potential difference. The friction electrode 4 and the friction medium layer accumulate equal amounts of positive and negative charges due to the triboelectric effect, such as Figure 6 As shown in ji;
[0088] (2) When the finger begins to approach the surface of the input device, driven by the potential difference, the charge flows from the friction electrode 4 to the friction electrode 3 through the external circuit. This current direction is defined as the forward current. When the piezoelectric sensing layer on the top is deformed due to external pressure, the positive and negative charges are separated inside the BaTiO3@PVDF material, and a voltage potential is generated between the upper and lower electrodes of the piezoelectric layer. The current flows from electrode 2 to electrode 1. This current direction is the same as the direction of the triboelectric current, as shown in Figure 6 As shown in j-ii;
[0089] (3) When the external force continues to be applied, the finger completely contacts the input device, the friction electrode 3 and the friction medium layer contact each other, and equal amounts of positive and negative charges are generated and distributed on the surface of the friction material. The piezoelectric deformation reaches the maximum state, that is, the stable state, such as Figure 6 As shown in j-iii;
[0090] (4) When the external force is removed, the finger starts to move away from the input device, and the top friction electrode 3 moves upward and separates from the friction medium layer. Driven by the potential difference, electrons flow from electrode 3 to electrode 4 through the external circuit. When the upper friction layer returns to its original position, the friction dielectric layer and electrode 4 accumulate equal and opposite charges, forming an electric field from bottom to top. When the piezoelectric layer rebounds to its original state, the internal polarization direction of the piezoelectric material changes, the polarity of the surface charge also changes, and the direction of the generated current changes, such as Figure 6 As shown in j-iv.
[0091] In order to build the AI model, a large number of keystroke dynamic samples and their behavioral characteristics were analyzed. For the biometric verification system, the consistency of keystroke dynamics is very important, but the behavioral information may change due to the user's psychological and physical condition. Therefore, the AI model needs to learn, adapt and implement the user's behavioral information and its impact on user identification and authentication;
[0092] In addition, the embodiment of the present invention also discloses a neural network model input device, such as Figure 7 Shown is a schematic diagram of a self-driven fusion sensing system based on keystroke dynamics using a neural network system for biometric password verification.
[0093] The neural network system uses 1DCNN as the classification model. For each training set, the convolutional neural network uses multiple convolution kernels to perform local convolution calculations on the input signal. By sharing the convolution kernel weight parameters, the number of parameters can be reduced and deeper features can be extracted. The convolution operation obtains linearly inseparable feature data, so an activation layer is added after the convolution layer to improve the linear separability of the model.
[0094] In addition, this article uses the activation function "ReLU", the gradient of which does not gradually decrease as the number of network layers increases, effectively solving the problem of gradient diffusion. After the input passes through the convolution layer, a large number of parameters will be generated, which contains some redundant information. Therefore, a pooling layer is added between the two convolution layers to reduce the parameters in the network through downsampling, while retaining important features, removing unimportant information and reducing the number of parameters.
[0095] In addition, the output of the last pooling layer is converted into a one-dimensional feature vector and then sent to the fully connected layer to pass the features extracted at the filtering stage to the classifier.
[0096] Figure 7 a shows the architecture of the biometric authentication system based on keystroke dynamics. The input device converts the keystroke mechanical energy into an electrical signal, which is then processed by the signal processing system and then extracted by the neural network model to classify and identify the waveform;
[0097] In addition, different key combinations can set different password waveforms. In this verification, a six-digit password (9-7-2-6-8-3) was set. In the experiment, four users entered the password "972683" 200 times each through the keystroke device;
[0098] In addition, among the 200 keystroke samples of each user, 100 random samples are used to construct the training set database, and the other 100 random data sets are used to construct the test set database.
[0099] The password waveform obtained by the fusion sensor is as follows Figure 7 As shown in d.
[0100] In order to distinguish the advantages of the hybrid sensor, the PENG sensor and TENG sensor were used separately to input the same password. The password waveforms are as follows: Figure 7 b and Figure 7 As shown in c.
[0101] We use the data from PENG sensors, TENG sensors, and fusion sensors to perform password and user identity recognition using the same neural network algorithm model, and the model accuracies obtained during training are shown in Figures 8e, 8f, and 8d, respectively.
[0102] In addition, users can log into the system only if they have established keystroke information, and the system can identify different users even if they enter the same code, such as Figure 7 An impostor who knows the correct password still cannot log into a computer system unless his / her keystroke dynamics match the information stored in the system, such as Figure 7 As shown by jb.
[0103] As another aspect of the present disclosure, a password and identity dual verification method based on keystroke dynamics is also provided, and the password and identity dual verification system based on keystroke dynamics is adopted, comprising:
[0104] Processing the digital signal to extract characteristic information, wherein the digital signal is the piezoelectric and frictional impedance input by the user tapping the input device converted by the wireless data acquisition and signal processing module; and
[0105] Use feature information to identify and verify users.
[0106] As another aspect of the present disclosure, a computer device is also provided, comprising:
[0107] Memory; and
[0108] The processor is configured to execute the password and identity dual verification method based on the instructions stored in the memory.
[0109] And, a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the instructions are executed by a processor, the password and identity dual verification method can be implemented.
[0110] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. An input device based on keystroke dynamics, characterized in that The invention comprises a piezoelectric sensor at the top and a triboelectric sensor array device at the bottom, wherein: The piezoelectric sensor uses BaTiO3@PVDF piezoelectric film as the piezoelectric layer; The triboelectric sensor uses a variety of materials as a triboelectric sensing array device, and the bottom electrode of the device uses a flexible interactive array electrode printed by FPCB.
2. The input device according to claim 1, characterized in that: Silver electrodes are deposited on both sides of the piezoelectric layer, and a PVDF film is used as a packaging layer. Silver electrodes are deposited on the surface of the bottom packaging layer of the piezoelectric sensor as a positive friction layer. PDMS is used as a spacer layer between the positive friction layer and the triboelectric sensor array device, and the size of the spacer layer is consistent with that of the flexible interactive array electrode.
3. The input device according to claim 1, characterized in that: The BaTiO3@PVDF piezoelectric film is prepared by the following method: S1. Preparation of primary crystallized film: BaTiO3 powder was added to DMF solution and ultrasonically stirred, and then PVDF-HFP powder was added and stirred and degassed to obtain a mixed solution; the mixed solution was cast on an aluminum plate and heated and dried to obtain a formed primary crystallized film; S2. Uniaxial stretching: The primary crystallized film is stretched at 58-62°C with a stretching rate of 5; S3. Thin film polarization: Use CM2400 epoxy conductive adhesive to stick the aluminum foil on the stretched film and solidify it. Then stick the electrode sheets on both sides of the film for oil bath polarization.
4. The input device according to claim 3, characterized in that: The oil bath polarization is as follows: the polarization electric field strength is 20 MV / m to 90 MV / m, the polarization strength is increased by 10 MV / m every 8 minutes, and the electric field strength is increased again after pausing for 4 minutes.
5. A password and identity dual verification system based on keystroke dynamics, characterized in that: include: Keystroke information collection module: comprising the input device based on keystroke dynamics as described in any one of claims 1 to 5, used to receive user keystrokes and input user behavior information and different waveform combinations, so as to realize the dual sensing functions of password and identity; Wireless data acquisition and signal processing module: coupled to the keystroke information acquisition module, used to receive the piezoelectric and frictional impedance output by the keystroke information acquisition module, fuse and add them, output them as digital signals, and transmit them to the PC serial port via wireless; Pattern recognition module: 1DNN neural network model, used to extract features of password signals from two dimensions: time domain and energy domain, then fuse them at the feature layer, and finally classify the signals to achieve user identification and password verification.
6. The verification system according to claim 5, characterized in that: The wireless data acquisition and signal processing module includes a fusion signal measurement circuit, which is composed of two high-impedance small voltage measurement circuits and a three-way adder, an AD sampling module, an STM32 microprocessor and a LoRa wireless transmission module; the high-impedance voltage measurement circuit is used to process the source impedance of piezoelectricity and triboelectricity, the three-way adder couples the piezoelectric and triboelectric signals, and the AD sampling module reads the coupled voltage signal as a digital signal; the STM32 microprocessor reads the AD signal at a frequency of 1KHz and transmits the data to the LoRa wireless transmission module, and the LoRa wireless transmission module receives the data at the PC end and transmits it to the PC serial port.
7. The verification system according to claim 5, characterized in that: The 1DCNN neural network model in the pattern recognition module is a four-layer network structure, including two convolutional layers and two fully connected layers. The convolutional layer uses the maximum pooling layer for sampling and the Same method is used for edge filling. The entire model is trained using an adaptive (Adam) optimizer.
8. A password and identity dual verification method based on keystroke dynamics, characterized in that: A password and identity dual verification system based on keystroke dynamics according to any one of claims 5 to 9, comprising: Piezoelectric and triboelectric sensor arrays, signal processing modules, and visual human-computer interfaces; user identification is performed using the user's keystroke signature information.
9. A computer device comprising: Memory; as well as A processor configured to execute the password and identity dual verification method as claimed in claim 8 based on the instructions stored in the memory.
10. A computer-readable storage medium storing computer instructions, wherein the instructions, when executed by a processor, can implement the password and identity dual verification method as claimed in claim 8.
