CN108287989B - A Trajectory-based Sliding Verification Code Human-Machine Recognition Method - Google Patents
A Trajectory-based Sliding Verification Code Human-Machine Recognition Method Download PDFInfo
- Publication number
- CN108287989B CN108287989B CN201810050045.8A CN201810050045A CN108287989B CN 108287989 B CN108287989 B CN 108287989B CN 201810050045 A CN201810050045 A CN 201810050045A CN 108287989 B CN108287989 B CN 108287989B
- Authority
- CN
- China
- Prior art keywords
- feature
- trajectory
- features
- model
- extract
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012795 verification Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 title claims abstract description 18
- 238000012549 training Methods 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000007477 logistic regression Methods 0.000 claims description 4
- 238000007637 random forest analysis Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 230000006399 behavior Effects 0.000 description 5
- 238000001514 detection method Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/36—User authentication by graphic or iconic representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Security & Cryptography (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
本发明公开了一种基于轨迹的滑动验证码人机识别方法，包括以下步骤：采集用户轨迹数据；根据轨迹数据构建多维特征体系；根据设计好的人机识别模型对多维特征体系进行轨迹区分。本发明中，通过结合人类轨迹的两种现象进行多维特征体系的设计，用特征描述用户的滑动验证码习惯，进而将用户操作与机器模仿进行区别开来，在与攻击者的黑产工具的对抗中能占据优势，起到较好的对抗保护作用。
The invention discloses a trajectory-based sliding verification code human-machine identification method, comprising the following steps: collecting user trajectory data; constructing a multi-dimensional feature system according to the trajectory data; In the present invention, a multi-dimensional feature system is designed by combining the two phenomena of human trajectory, and the user's sliding verification code habit is described by features, and then the user operation is distinguished from the machine imitation. In confrontation, it can take advantage and play a better role in confrontation and protection.
Description
Technical Field
The invention relates to the technical field of biometric authentication, in particular to a track-based sliding verification code man-machine identification method.
Background
The sliding verification code serving as a biological authentication technology can meet the requirement of the current network environment on identity authentication security, and is widely applied to various man-machine verification products. Meanwhile, the method is also focused by attackers, and the attackers develop black products capable of simulating human behaviors to challenge mouse tracks in the sliding verification code verification process.
The attacker generates a humanoid trace batch operation through the blackout facility to bypass detection, and continuously upgrades its counterfeit data during the countermeasure process to continuously bypass the same upgraded detection technique. The existing detection technology mainly aims at machine identification, a mode of countermeasures against continuously updated machine behaviors has hysteresis, and detection and update are always performed after a black production tool causes certain loss. Therefore, in the technical countermeasures with both sides being upgraded, it is important how to take advantage of the countermeasures with the black production tools of the attackers.
Disclosure of Invention
The invention aims to: the track-based sliding verification code man-machine identification method aims to construct a multi-dimensional effective characteristic system to identify a trigger of a sliding verification code and ensure the safety of a network environment protected by verification behaviors.
The technical scheme adopted by the invention is as follows:
a track-based sliding verification code man-machine identification method comprises the following steps:
s1: collecting user track data;
s2: constructing a multi-dimensional characteristic system according to the track data;
s3: and distinguishing the tracks of the multi-dimensional characteristic system according to the designed human-computer recognition model.
Further, the multi-dimensional feature system comprises an X feature, a Y feature and a T feature.
Further, the specific steps of the X feature extraction are as follows:
s201: extracting an X characteristic class, and carrying out normalization processing on a track transverse coordinate X;
s202: dividing the track transverse coordinate into a front half section and a rear half section;
s203: respectively extracting a plurality of X characteristic groups of a front half section X front, a rear half section X rear, a front half section collar difference X front diff, a rear half section collar difference X rear diff and a stop section final stop of a track;
s204: and extracting features in each X feature group, including maximum value, peak value, median value, variance, minimum value and range.
Further, the Y feature extraction specifically comprises the following steps:
s211: extracting a Y characteristic class, and carrying out normalization processing on the track longitudinal coordinate Y;
s212: respectively extracting a plurality of Y characteristic groups of the whole section Y, the half-folded Y half, the whole section adjacent potential difference Y diff and the whole section adjacent potential difference Y diff;
s213: features in each Y feature group are extracted, including variance, mean, range, and sum.
Further, the specific steps of the T feature extraction are as follows:
s221: extracting a T characteristic class, and carrying out normalization processing on the time characteristic T;
s222: extracting a T-X characteristic group, and subtracting the normalized time characteristic T from the normalized transverse coordinate X;
s223: and extracting features in the T-X feature group, including maximum value, peak value, median value, variance, minimum value and range.
Further, the step S3 is a specific step of designing a human-machine recognition model as follows:
s301: inputting the characteristics in the multi-dimensional characteristic system into a plurality of training models for algorithm training;
s302: and carrying out linear weighting on the training output of the feature algorithm.
Further, the training model comprises: a Catboost model, an XGboost model, a Randomforest model and a Logistic regression model.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
1. in the invention, a multi-dimensional characteristic system is designed by combining two phenomena of human tracks, the sliding verification habit of a user is described by using characteristics, and further the user operation is distinguished from machine simulation, so that the system has advantages in the countermeasures with black production tools of attackers, and plays a good countermeasure protection role.
2. In the invention, the behavior habit of the human during the sliding verification code is described by mainly adopting the transverse characteristic x, the characteristic of the machine is described by using the longitudinal characteristic y, and the difference between the human and the machine is described by using the time characteristic t as a supplement, so that the user operation and the machine simulation can be more accurately distinguished, and the accuracy of track distinguishing is improved.
3. In the invention, the actual verification effect is on a test set recorded by 200 ten thousand tracks, the harmonic F value of the accuracy rate and the recall rate reaches 88.56, and is far higher than the effect 87.89 of the scheme mainly describing a 'machine'.
Drawings
FIG. 1 is a graph of a multi-dimensional feature system feature relationship of the present invention;
FIG. 2 is a conceptual diagram of a multi-dimensional feature hierarchy of the present invention;
FIG. 3 is a diagram of a human-machine recognition model relationship according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Example 1
A track-based sliding verification code man-machine identification method comprises the following steps:
s1: collecting user track data;
the method comprises the steps of collecting user track data (x, y, t), including transverse coordinates x and longitudinal coordinates y of different time points t in the track triggering process, specifically, obtaining track records of a user in the sliding verification code triggering process, and providing data support for the construction of a sliding verification code multi-dimensional feature system.
S2: constructing a multi-dimensional characteristic system according to the track data;
based on the discovery of two modes, one, the terminal fold-back phenomenon of the human trajectory; secondly, the phenomenon of the human track is far, fast and slow; and performing multi-dimensional feature extraction to further construct a multi-dimensional feature system, so that later algorithm training of a man-machine recognition model is facilitated, and a difference between the man and the machine is found by using feature description instead of using feature description of the machine.
S3: carrying out track distinguishing on a multi-dimensional characteristic system according to a designed man-machine recognition model;
and inputting the characteristics of the multi-dimensional characteristic system into a designed man-machine recognition model for man-machine recognition model optimization, inputting the characteristics of the multi-dimensional characteristic body into the man-machine recognition model for feature learning to obtain different probability values, and enabling the probability output by the man-machine recognition model to be closer to the real type of the track in a linear weighting mode.
In the invention, the human track is analyzed, and based on the discovery of two modes, the mode I is as follows: human trajectory retrace phenomenon, mode two: the method comprises the steps of constructing a multi-dimensional characteristic system for the phenomena of far, fast and near human tracks, further designing a human-computer recognition model, as shown in fig. 1, constructing the multi-dimensional characteristic system for a training set corresponding to user tracks, further training the human-computer recognition model, constructing the same multi-dimensional characteristic system for a prediction set corresponding to predicted tracks, inputting the multi-dimensional characteristic system into the trained human-computer recognition model for model optimization and track type distinguishing training, and performing linear weighting on training output to obtain a probability value for distinguishing track types.
Example 2
On the basis of the embodiment 1, the multi-dimensional feature system comprises X features, Y features and T features.
The invention mainly adopts the transverse characteristic x to describe the behavior habit of the human when performing the sliding verification, uses the longitudinal characteristic y to describe the characteristic of the machine, and uses the time characteristic T as a supplement to describe the difference between the human and the machine, as shown in figure 2.
Further, the specific steps of the X feature extraction are as follows:
s201: extracting an X characteristic class, and carrying out normalization processing on a track transverse coordinate X;
s202: the track is divided into a first half section and a second half section by combining the phenomenon of 'far quick and near slow' of the human track mode II;
specifically, "far and near" indicates that the speed is faster when the verification code is slid farther from the target point and slower when the verification code is slid closer to the target point. Therefore, the horizontal x-feature group constructively divides the track into the first and second half segments to be extracted respectively.
S203: respectively extracting a plurality of X characteristic groups of a front half section X front, a rear half section X rear, a front half section collar difference X front diff, a rear half section collar difference X rear diff and a stop section final stop of a track;
and combining the phenomenon of 'end retrace' of the human track mode one to construct an extraction stop final stop feature group.
Specifically, the transverse coordinate data of the whole track extracted from the track data (x, y, t) form a transverse sequence { x }1,x2,...,xt,...,xnGet the first half of the trace sequence { x }1,x2,...,xn/2Form x _ front, take the second half of the trace sequence { x }n/2,xn/2+1,...,xnConstitute x _ real, take the difference of the ortho position of the first half of the trajectory sequence { x }2-x1,x3-x2,.. constitute x _ front _ diff, taking the difference of the neighbors of the second half of the trajectory sequencen-1-xn-2,xn-xn-1And (5) forming an x _ real _ diff, and combining a human track mode one, and taking the last fifth of the track sequence to form a stop segment final _ stop.
S204: and extracting features in each X feature group, including maximum value, peak value, median value, variance, minimum value and range.
According to the invention, the transverse features of the feature system are designed from multiple dimensions, so that the input can be better provided for the model.
Further, the Y feature acquisition specifically includes the following steps:
s211: extracting a Y characteristic class, and carrying out normalization processing on the track longitudinal coordinate Y;
specifically, longitudinal coordinate data of the whole track extracted from the track data (x, y, t) form a transverse sequence { y }1,y2,...,yt,...,ynAnd normalizing the sequence.
S212: respectively extracting a plurality of Y characteristic groups of the whole section Y, the half-folded Y half, the whole section adjacent potential difference Y diff and the whole section adjacent potential difference Y diff;
specifically, for the normalized vertical sequence y1,y2,...,yt,...,ynGet the whole segment { y in the trace sequence1,y2,...,yt,...,ynForm y, take the whole segment in the trace sequence y1,y2,...,yt,...,ynSubtracting 0.5 from the sequence to form y _ half, and taking the adjacent position difference of the whole track sequence (y)2-y1,y3-y2,. } forming y _ diff, taking the neighboring difference of neighboring differences of the whole segment in the trajectory sequence { (y)3-y2)-(y2-y1) ,. constitute y _ diffdiffdiff.
S213: features in each Y feature group are extracted, including variance, mean, range, and sum.
When the transverse feature X is less descriptive in a certain track, the longitudinal feature y which is better descriptive to the 'machine' can play an auxiliary role in the model discrimination.
Further, the specific steps of the T-feature acquisition are as follows:
s221: extracting a T characteristic class, and carrying out normalization processing on the time characteristic T;
specifically, the independent time series are only sampling marks, have no good interpretability, but have better expression significance when combined with the transverse x series.
And (3) representing the speed condition in the track generation process on another layer by utilizing the normalized X minus T, and providing better characteristic input for the model.
S222: extracting a T-X characteristic group, and subtracting the normalized time characteristic T from the normalized transverse coordinate X;
specifically, the transverse coordinate data of the whole track extracted from the track data (x, y, t) form a transverse sequence { x }1,x2,...,xt,...,xnAnd time series t1,t2,...,tnAre normalized separately, and then the transverse sequence { x1,x2,...,xt,...,xnSubtract the time series t1,t2,...,tn}。
S223: and extracting features in the T-X feature group, including maximum value, peak value, median value, variance, minimum value and range.
In summary, the feature groups and feature lists provided in this embodiment are shown in table 1:
example 3
On the basis of the embodiment 1, the step S3 is a specific step of designing a human-machine recognition model as follows:
s301: respectively inputting the characteristics in the multi-dimensional characteristic system into a plurality of training models for algorithm training;
s302: carrying out linear weighting on the training output of the feature algorithm;
further, the training model comprises: a Catboost model, an XGboost model, a Randomforest model and a Logistic regression model.
As shown in fig. 3, specifically, probability values output by training of the castboost model, the XGBoost model, the RandomForest model, and the logistic regression model are linearly weighted, so as to obtain the human-computer recognition model linearly weighted by the four basic models.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (3)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810050045.8A CN108287989B (en) | 2018-01-18 | 2018-01-18 | A Trajectory-based Sliding Verification Code Human-Machine Recognition Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810050045.8A CN108287989B (en) | 2018-01-18 | 2018-01-18 | A Trajectory-based Sliding Verification Code Human-Machine Recognition Method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108287989A CN108287989A (en) | 2018-07-17 |
| CN108287989B true CN108287989B (en) | 2021-05-28 |
Family
ID=62835600
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810050045.8A Active CN108287989B (en) | 2018-01-18 | 2018-01-18 | A Trajectory-based Sliding Verification Code Human-Machine Recognition Method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108287989B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109033793A (en) * | 2018-07-11 | 2018-12-18 | 北京京东金融科技控股有限公司 | Slide the man-machine recognition methods of identifying code and device, electronic equipment and readable medium |
| CN109241709B (en) * | 2018-08-03 | 2023-04-07 | 平安科技（深圳）有限公司 | User behavior identification method and device based on slider verification code verification |
| CN110889103B (en) * | 2018-09-07 | 2024-04-05 | 京东科技控股股份有限公司 | Method and system for verifying sliding block and model training method thereof |
| CN109086594A (en) * | 2018-09-21 | 2018-12-25 | 郑州云海信息技术有限公司 | A kind of method, apparatus and system for realizing the verifying of sliding-type identifying code |
| CN109543393A (en) * | 2018-10-23 | 2019-03-29 | 武汉极意网络科技有限公司 | Behavior verification method and device |
| CN109522692B (en) * | 2018-11-19 | 2019-10-18 | 第四范式（北京）技术有限公司 | Webpage machine behavioral value method and system |
| CN109600361B (en) * | 2018-11-26 | 2021-05-04 | 武汉极意网络科技有限公司 | Hash algorithm-based verification code anti-attack method and device, electronic equipment and non-transitory computer readable storage medium |
| CN109726543B (en) * | 2018-12-30 | 2021-10-08 | 北京城市网邻信息技术有限公司 | Login method and device of application program, terminal equipment and storage medium |
| CN110046647A (en) * | 2019-03-08 | 2019-07-23 | 同盾控股有限公司 | A kind of identifying code machine Activity recognition method and device |
| CN110807183A (en) * | 2019-10-12 | 2020-02-18 | 广州多益网络股份有限公司 | Sliding verification code man-machine behavior identification method of multi-dimensional feature system |
| CN111310156B (en) * | 2019-11-28 | 2022-08-19 | 苏宁金融科技(南京)有限公司 | Automatic identification method and system for slider verification code |
| CN111666968A (en) * | 2020-04-21 | 2020-09-15 | 北京奇艺世纪科技有限公司 | Man-machine recognition method and device, electronic equipment and computer readable storage medium |
| CN114117390A (en) * | 2021-11-12 | 2022-03-01 | 东方财富信息股份有限公司 | Machine learning and interactive man-machine distinguishing verification generation method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975823A (en) * | 2016-05-05 | 2016-09-28 | 百度在线网络技术（北京）有限公司 | Verification method and apparatus used for distinguishing man and machine |
| US9483292B2 (en) * | 2010-11-29 | 2016-11-01 | Biocatch Ltd. | Method, device, and system of differentiating between virtual machine and non-virtualized device |
| CN106155298A (en) * | 2015-04-21 | 2016-11-23 | 阿里巴巴集团控股有限公司 | Man-machine recognition methods and device, the acquisition method of behavior characteristics data and device |
| CN106815514A (en) * | 2015-11-27 | 2017-06-09 | 阿里巴巴集团控股有限公司 | A kind of man-machine recognition methods and device |
-
2018
- 2018-01-18 CN CN201810050045.8A patent/CN108287989B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9483292B2 (en) * | 2010-11-29 | 2016-11-01 | Biocatch Ltd. | Method, device, and system of differentiating between virtual machine and non-virtualized device |
| CN106155298A (en) * | 2015-04-21 | 2016-11-23 | 阿里巴巴集团控股有限公司 | Man-machine recognition methods and device, the acquisition method of behavior characteristics data and device |
| CN106815514A (en) * | 2015-11-27 | 2017-06-09 | 阿里巴巴集团控股有限公司 | A kind of man-machine recognition methods and device |
| CN105975823A (en) * | 2016-05-05 | 2016-09-28 | 百度在线网络技术（北京）有限公司 | Verification method and apparatus used for distinguishing man and machine |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108287989A (en) | 2018-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108287989B (en) | A Trajectory-based Sliding Verification Code Human-Machine Recognition Method | |
| Eberle et al. | Building and interpreting deep similarity models | |
| CN104809377B (en) | Network user identity monitoring method based on webpage input behavior feature | |
| CN104123500B (en) | A kind of Android platform malicious application detection method and device based on deep learning | |
| Wang et al. | Kinect based dynamic hand gesture recognition algorithm research | |
| Harouni et al. | Online Persian/Arabic script classification without contextual information | |
| Xia et al. | Enhancing backdoor attacks with multi-level mmd regularization | |
| CN107766852A (en) | A kind of man-machine mouse track detection method based on convolutional neural networks | |
| Ge et al. | Semantic-guided reinforced region embedding for generalized zero-shot learning | |
| Li et al. | Weakly-supervised 3d spatial reasoning for text-based visual question answering | |
| Ruan et al. | Dynamic gesture recognition based on improved DTW algorithm | |
| CN115100739B (en) | Man-machine behavior detection method, system, terminal device and storage medium | |
| CN105989287A (en) | Method and system for judging homology of massive malicious samples | |
| CN105303163A (en) | Method and detection device for target detection | |
| CN117237688A (en) | A method for tracing the origin of APT malicious samples attacking organizations based on fusion features | |
| CN115865458B (en) | Network attack behavior detection method, system and terminal based on LSTM and GAT algorithm | |
| Tong et al. | 1D-DRSETL: a novel unsupervised transfer learning method for cross-condition fault diagnosis of rolling bearing | |
| Zhang et al. | Deep-learning model extraction through software-based power side-channel | |
| Mahfouz et al. | A behavioral biometric authentication framework on smartphones | |
| Rahnama et al. | An adversarial approach for explaining the predictions of deep neural networks | |
| CN107077617A (en) | fingerprint extraction method and device | |
| Sharif et al. | Function identification in android binaries with deep learning | |
| CN109857783B (en) | Sonar seepage motion law discovery method based on deep learning | |
| Sarma et al. | Handwritten Assamese numeral recognizer using HMM & SVM classifiers | |
| Xie et al. | Driving pattern recognition based on improved LDA model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211228 Address after: 908, block a, floor 8, No. 116, Zizhuyuan Road, Haidian District, Beijing 100089 Patentee after: ZHONGZI DATA CO.,LTD. Patentee after: Zhongzi highway maintenance and Inspection Technology Co., Ltd Address before: No. 30 Xueyuan Road, Haidian District, Beijing 100083 Patentee before: University OF SCIENCE AND TECHNOLOGY BEIJING |