CN115225383A - Online Web site defense method based on countermeasure patch - Google Patents
Online Web site defense method based on countermeasure patch Download PDFInfo
- Publication number
- CN115225383A CN115225383A CN202210852629.3A CN202210852629A CN115225383A CN 115225383 A CN115225383 A CN 115225383A CN 202210852629 A CN202210852629 A CN 202210852629A CN 115225383 A CN115225383 A CN 115225383A
- Authority
- CN
- China
- Prior art keywords
- patch
- countermeasure
- defense
- padding
- web site
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000007123 defense Effects 0.000 title claims abstract description 34
- 238000002347 injection Methods 0.000 claims abstract description 32
- 239000007924 injection Substances 0.000 claims abstract description 32
- 230000008569 process Effects 0.000 claims abstract description 19
- 238000004891 communication Methods 0.000 claims abstract description 3
- 238000010586 diagram Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 12
- 238000005457 optimization Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 101150064138 MAP1 gene Proteins 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 101150077939 mapA gene Proteins 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/04—Network architectures or network communication protocols for network security for providing a confidential data exchange among entities communicating through data packet networks
- H04L63/0407—Network architectures or network communication protocols for network security for providing a confidential data exchange among entities communicating through data packet networks wherein the identity of one or more communicating identities is hidden
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/04—Network architectures or network communication protocols for network security for providing a confidential data exchange among entities communicating through data packet networks
- H04L63/0407—Network architectures or network communication protocols for network security for providing a confidential data exchange among entities communicating through data packet networks wherein the identity of one or more communicating identities is hidden
- H04L63/0421—Anonymous communication, i.e. the party's identifiers are hidden from the other party or parties, e.g. using an anonymizer
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
- H04L63/1416—Event detection, e.g. attack signature detection
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
- H04L63/1425—Traffic logging, e.g. anomaly detection
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1441—Countermeasures against malicious traffic
- H04L63/1466—Active attacks involving interception, injection, modification, spoofing of data unit addresses, e.g. hijacking, packet injection or TCP sequence number attacks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/02—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses an online Web site defense method based on a countermeasure patch, which comprises the steps of firstly calculating a key area of network flow data; and generating an anti-patch: firstly, analyzing the difference between online and offline defense scenes, limiting the generation process of a countermeasure patch, then providing a key area-oriented non-target pertinence countermeasure patch generation algorithm, and generating a countermeasure patch corresponding to each site; and finally, injecting an anti-patch: the injection process is based on a WFPadTools framework to design a Tor plug-in, firstly, plug-in deployment is carried out at a Tor user side and a bridge side, then, a control message is designed for communication between the user side and the bridge side, and finally, the countermeasure patch is injected online according to a set strategy. The invention improves the practicability of the fingerprint defense of the Web site by utilizing the anti-patch technology and designing the patch injection scheme.
Description
Technical Field
The invention relates to a Web site fingerprint defense technology, belongs to the technical field of Traffic Analysis (Traffic Analysis) and anonymous Network (Anonymity Network), and particularly relates to an online Web site defense method based on a countermeasure patch.
Background
The continuous development of the internet has led to continuous innovation of Web technologies, and the Web carries an increasingly large number of service types, which also contain a large amount of user privacy data. The Web site fingerprint attack is a technology for extracting target site features by using a channel measurement analysis method so as to identify sites, firstly, an anonymous communication system is used for simulating a user to access a target supervision site to obtain flow, and a fingerprint is formed and a model is established for the obtained flow extraction mode features; and then extracting the same characteristics from the anonymous traffic generated by the actual access of the user to identify so as to presume the actual site accessed by the user. Therefore, the research on the defense technology is of great significance to the protection of the privacy of the user. However, existing defense technologies for deep learning of the Web site fingerprint attack are all in an offline scene, and a defender needs to generate a theoretical optimization defense strategy for a single flow after obtaining a flow data set, so that the defense method cannot be effectively deployed in a real environment, and the practicability of the Web site fingerprint defense method is greatly reduced.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problem that the practicability of a Web site fingerprint defense technology is poor in a real scene, the invention provides an online Web site defense method based on a countermeasure patch.
The invention adopts the following technical scheme:
an online Web site defense method based on countermeasure patches comprises the following steps:
(1) Determining a key area: dividing the network traffic sequence of the station into a plurality of areas, and calculating the contribution degree of each area to the classification result, wherein the area with higher contribution degree is called a key area. Determining the critical area may help defenders find a good injection location.
(2) Anti-patch generation: analyzing the difference between online and offline defense scenes, limiting the generation process of the countermeasure patch, and generating a key region-oriented non-target countermeasure patch
(3) Anti-patch injection: designing a Tor plug-in based on a WFPadTools framework, deploying at a Tor client and a bridge client, and injecting the anti-patch online according to an injection position l.
Further, the step (1) specifically includes:
(11) Inputting the network flow sequence into the convolutional neural network and obtaining the output A of the last convolutional layer and the target label score y c (i.e., without the value of softmax passed). Wherein A contains Z characteristic maps in total, and the size of each characteristic map is Z =1*I. The value of the ith column in the kth feature map is
(12) Calculating the gradient of each value in the characteristic diagram by using back propagation
As shown in equation (1), and the gradient of the feature map
Averaging to obtain the weight of the feature map
As shown in equation (2):
(13) Obtaining the weight of the characteristic diagram, and then obtaining the value A of each characteristic diagram k And weight
Adding the multiplied values to obtain the contribution degree f of each point c The calculation process is shown in formula (3):
(14) Degree of contribution f c Performing ReLU operation to obtain L c Is prepared by mixing L c The data are divided into a plurality of groups according to the length 50, and the contribution degrees of each element in the groups are added to obtain the group contribution degree.
Further, the step (2) specifically includes:
(21) Limiting patch elements to be +1 or-1, splitting the patch into a plurality of small patches, setting the first element of each small patch to be-1 and the last element to be +1;
(22) Calculating the number N of the small patches according to the length L of the network flow sequence, the bandwidth load percentage B and the small patch length plen, as shown in a formula (4):
(23) Obtaining a flow key area through the step (1), and selecting a key area with high front Q contribution degree of the front group to obtain an injection position l;
(24) The countermeasure patch is generated as shown in equation (5):
wherein X is a source site traffic data set, s represents an injection position, psi represents a patch injection mode of network traffic, and y t Representing the site label, F representing the fingerprint classifier of the Web site, and the loss function uses the cross entropy of the crossEntropyLoss class, and updates the patch by maximizing the classification loss of the fingerprint classifier.
Further, the step (3) specifically includes:
(31) When a user initiates a site access request, a client records the directions of all data packets to generate a real-time data packet sequence;
(32) When the flow sequence reaches the injection position, the client loads the small patch at the position, caches the received uplink data packet into the Buffer, and sends a PADDING _ START control message and the number G of the filling messages to the bridge end;
(33) After receiving the PADDING _ START message, the bridge end firstly sends G-1 PADDING data messages, and finally sends a PADDING _ STARTED control message to inform the client of the START of filling;
(34) Alternately sending the PADDING data message according to the small patch strategy;
(35) After the small patch injection is finished, the client sends a PADDING _ END message and sends a data packet cached in the Buffer;
(36) After receiving the PADDING _ END message, the network bridge END sends a data packet cached in the Buffer;
(37) And repeating the processes (31) to (36), and injecting each confrontation small patch according to the confrontation patch strategy.
Has the advantages that: compared with the prior art, the invention has the following remarkable advantages:
1. and determining a key area of the flow, finding a position more suitable for inserting a patch, and reducing bandwidth overhead and practice overhead brought by defense while improving the success rate of the defense.
2. Most Web site fingerprint defense schemes generate a theoretical optimized defense strategy aiming at single flow, and the invention generates a general countermeasure patch for each site by utilizing a countermeasure patch technology, can adapt to network flow change generated by problems such as network delay and the like, and improves the practicability of the Web site fingerprint defense technology.
3. The difference between the online defense scene and the offline defense scene is analyzed, and the generation process of the anti-patch is limited, so that the generated defense strategy better meets the requirements of the online scene, and the practicability of the Web site defense technology is improved.
Drawings
FIG. 1 is a schematic diagram of the defense scenario of the present invention.
Fig. 2 is a schematic diagram of determining a key area according to the present invention.
Fig. 3 is a schematic diagram of the anti-patch injection process of the present invention.
Detailed Description
The invention designs and realizes an online Web site fingerprint defense technology based on a countermeasure patch, which is divided into three parts, namely key area determination, countermeasure patch generation and countermeasure patch injection. The method specifically comprises the following steps:
key zone determination
Dividing the network traffic sequence of the station into a plurality of areas, and calculating the contribution degree of each area to the classification result, wherein the area with higher contribution degree is called a key area. Determining a key region can help defenders to find a good injection position and destroy a flow sequence in the region, and the key region determination process is shown in fig. 2:
inputting the network flow sequence into the convolutional neural network and obtaining the output A of the last convolutional layer and the target label score y c (i.e., without the value of softmax passed). Wherein A contains Z feature maps, and the size of each feature map is Z =1*I. The value of the ith column in the kth feature map is
Calculating the gradient of each value in the feature map by using back propagation
As shown in equation (1), and the gradient of the feature map
Averaging to obtain weights of feature maps
As shown in equation (2). Obtaining the weight of the characteristic diagram, and then obtaining the value A of each characteristic diagram k And weight
Adding the multiplied values to obtain the contribution degree f of each point c The calculation process is shown in formula (3).
Degree of contribution f c Performing ReLU operation to obtain L c Is prepared by mixing L c The data are divided into a plurality of groups according to the length 50, and the contribution degrees of each element in the groups are added to obtain the group contribution degree. The specific algorithm flow is as follows:
resist patch generation
The countermeasure patch technology is derived from the field of computer vision, belongs to a method for injecting countermeasure disturbance, but different from the traditional countermeasure disturbance method, a defender is not limited to injecting tiny disturbance which is specific to an original sample and is difficult to detect, but places a target on a general countermeasure patch, and the classifier can output wrong target classes for generating good patches no matter the patches are placed at any positions of any samples. The scheme analyzes the difference between the online defense scene and the offline defense scene, limits the generation process of the countermeasure patch, and generates a non-target pertinence countermeasure patch facing to a key area.
The invention limits the generation process of the traditional countermeasure patch, so that the traditional countermeasure patch is better adapted to the online Web site fingerprint defense scene, and the specific algorithm flow is as follows:
firstly, calculating a contribution score Socre of a source station group through (1), then calculating the minimum length L of a station traffic sequence in a source station traffic data set, and setting an original station traffic data set X w The length p _ len of each section of small patch, the bandwidth load percentage B, the iteration times T, the classification model M and the Loss function Loss are used as parameters and input into the algorithm.
The algorithm firstly obtains the maximum number of the filling packets which can be injected by calculating L & ltB & gt, and then divides the result by plen to obtain the number N of the small patches, namely the small patches are required to be injected into N groups, and the size of N does not exceed the total number of the groups. And then, initializing N small patches by the algorithm to form a complete confrontation patch, and selecting a group with high contribution degree front Q from the group contribution scores to obtain an injection position S. Next, injecting a countermeasure patch into each original network traffic sequence according to the injection position S, specifically, for each key region, randomly selecting an injection position in the region by the algorithm to perform small patch injection, because the patch cannot be guaranteed to be accurately injected to the specified position in the online traffic injection process, and therefore, the randomly selecting a position can guarantee that the patch has robustness to the change of the injection position inside the region.
Thereafter, the algorithm calculates the loss value after patch injection and updates the patch element values by back-propagation using Stochastic Gradient Descent (SGD). To apply the SGD, we relax the constraint on the elements within the patch during the patch optimization, allowing its value to be a continuous value between [ -1,1], and after the optimization ends we turn it to-1 for elements less than 0 within the patch and +1 for elements greater than 0 within the patch. This approach is widely used to solve the discrete optimization problem and ensure smooth back propagation. And finally, iterating for T times to generate a countermeasure patch which is universal to the site.
Resist patch injection
And the Tor user side and the Tor bridge side are required to be matched with each other for the anti-patch injection, and the generated anti-patch sequence is injected into the flow sequence in real time according to the patch position. Therefore, the Tor plug-in is written by means of the WFPadTools open source framework to realize network packet delay and injection of virtual packets.
The present invention uses modifications to the torrc file to enable Tor to use custom plug-ins and bridges for network traffic transport. The core configuration is as follows:
client configuration
Bridge end configuration
AssumeReachable 1
PublishServerDescriptor 0
Exitpolicy reject*:*
ORPort 4053
BridgeRelay 1
ServerTransportListenAddrMLPatch 0.0.0.0:40535
ServerTransportPluginMLPatch exec/wfpadtools/bin/obfsproxy\
--log-min-severity＝debug\
--log-file＝/root/mlpatch_server/mlpatch-server.log\
managed
As shown in fig. 3, the anti-patch injection process is performed in two states, namely a normal state and a filling state, where in the normal state, the two ends do not delay and fill the message, and only forward the received message, and in the filling state, the two ends store the received message in a Buffer and send the filling message to each other. When a user initiates a site access request, the client records the direction of each data packet and generates a real-time data packet sequence. Once the flow sequence reaches the filling point, the client loads the countermeasure patch firstly, switches the state of the client to the filling state, caches the received uplink data packet into the Buffer, and finally sends a PADDING _ START control message and the quantity G of the filling message to the bridge end. After receiving the PADDING _ START message, the bridge end first converts the state of the bridge end into a filling state, then sends G-1 PADDING data messages, and the last message sends PADDING _ STARTED to inform the client that filling is STARTED and the first segment of message is sent completely. After receiving the PADDING _ STARTED message, the client side continues to send the PADDING message according to the anti-patch strategy. And finally, the client sends a PADDING _ END message until the patch filling is finished, switches the state of the client into a normal state and sends the data packet cached in the Buffer. After receiving the PADDING _ END message, the network bridge END also switches the state to a normal state and sends the data packet cached in the Buffer. And completing the injection of one small patch, and repeating the operations for each small patch in the patch to finally complete the injection of the countermeasure patch.
Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the present application and not for limiting the scope of protection thereof, and although the present application is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: numerous variations, modifications, and equivalents will occur to those skilled in the art upon reading the present application and are within the scope of the claims appended hereto.
Claims (4)
1. An online Web site defense method based on countermeasure patches is characterized by comprising the following steps:
(1) Calculating a key area of the network traffic data: in the key area calculation process, a network flow sequence of a site is divided into a plurality of areas, and the group contribution degree of each area to a classification result is calculated, wherein the area with higher group contribution degree is called a key area;
(2) Generating a countermeasure patch: firstly, analyzing the difference between online and offline defense scenes, limiting the generation process of a countermeasure patch, then providing a key area-oriented non-target pertinence countermeasure patch generation algorithm, and generating a countermeasure patch corresponding to each site;
(3) The countermeasure patch is injected: the injection process is based on a WFPadTools framework to design a Tor plug-in, firstly, plug-in deployment is carried out at a Tor user side and a bridge side, then, a control message is designed for communication between the user side and the bridge side, and finally, the countermeasure patch is injected online according to a set strategy.
2. The method of claim 1, wherein the online Web site defense based on countermeasure patches comprises: the specific process of the step (1) is as follows:
(11) Inputting the network flow sequence into the convolutional neural network and obtaining the output A of the last convolutional layer and the target label score y c (ii) a Wherein A contains Z feature maps in total, the size of each feature map is Z =1*I, and the value of the ith column in the kth feature map is
(12) Calculating the gradient of each value in the feature map by using back propagation
As shown in equation (1), and the gradient of the feature map
Averaging to obtain weights of feature maps
As shown in equation (2):
(13) Obtaining the weight of the characteristic diagram, and then obtaining the value A of each characteristic diagram k And weight
Adding the multiplied values to obtain the contribution degree f of each point c The calculation process is shown in formula (3):
(14) Degree of contribution f c Performing ReLU operation to obtain L c Is prepared by mixing L c The data are divided into a plurality of groups according to the length 50, and the contribution degrees of each element in the groups are added to obtain the group contribution degree.
3. The method of claim 1, wherein the online Web site defense based on countermeasure patches comprises: the step (2) specifically comprises:
(21) Limiting patch elements to be +1 or-1, splitting the patch into a plurality of small patches, setting the first element of each small patch to be-1 and the last element to be +1;
(22) Calculating the number N of the small patches according to the length L of the network flow sequence, the bandwidth load percentage B and the small patch length plen, as shown in a formula (4):
(23) Obtaining a flow key area through the step (1), and selecting a key area with high front Q contribution degree of the front group to obtain an injection position l;
(24) The countermeasure patch is generated as shown in equation (5):
wherein X is a source site traffic data set, s represents an injection position, psi represents a patch injection mode of network traffic, and y t Representing a site label, F representing a fingerprint classifier of the Web site, and a loss function uses cross entropy of a cross EntropyLoss class, and updates the patch by maximizing the classification loss of the fingerprint classifier.
4. The method of claim 1, wherein the method comprises: the step (3) specifically comprises:
(31) When a user initiates a site access request, a client records the directions of all data packets to generate a real-time data packet sequence;
(32) When the flow sequence reaches the injection position, the client loads the small patch at the position, caches the received uplink data packet into the Buffer, and sends a PADDING _ START control message and the number G of the filling messages to the bridge end;
(33) After receiving the PADDING _ START message, the bridge end firstly sends G-1 PADDING data messages, and finally sends a PADDING _ STARTED control message to inform the client of the START of filling;
(34) Alternately sending the PADDING data message according to the small patch strategy;
(35) After the small patch injection is finished, the client sends a PADDING _ END message and sends a data packet cached in the Buffer;
(36) After receiving the PADDING _ END message, the network bridge END sends a data packet cached in the Buffer;
(37) And repeating the processes (31) to (36), and injecting each confrontation small patch according to the confrontation patch strategy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210852629.3A CN115225383A (en) | 2022-07-20 | 2022-07-20 | Online Web site defense method based on countermeasure patch |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210852629.3A CN115225383A (en) | 2022-07-20 | 2022-07-20 | Online Web site defense method based on countermeasure patch |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115225383A true CN115225383A (en) | 2022-10-21 |
Family
ID=83612635
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210852629.3A Pending CN115225383A (en) | 2022-07-20 | 2022-07-20 | Online Web site defense method based on countermeasure patch |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115225383A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180026994A1 (en) * | 2016-07-22 | 2018-01-25 | Alibaba Group Holding Limited | Network attack defense system and method |
| CN112241790A (en) * | 2020-12-16 | 2021-01-19 | 北京智源人工智能研究院 | Small countermeasure patch generation method and device |
| CN113938291A (en) * | 2021-09-03 | 2022-01-14 | 华中科技大学 | Encrypted flow analysis defense method and system based on anti-attack algorithm |
-
2022
- 2022-07-20 CN CN202210852629.3A patent/CN115225383A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180026994A1 (en) * | 2016-07-22 | 2018-01-25 | Alibaba Group Holding Limited | Network attack defense system and method |
| CN112241790A (en) * | 2020-12-16 | 2021-01-19 | 北京智源人工智能研究院 | Small countermeasure patch generation method and device |
| CN113938291A (en) * | 2021-09-03 | 2022-01-14 | 华中科技大学 | Encrypted flow analysis defense method and system based on anti-attack algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 杨弋鋆;邵文泽;王力谦;葛琦;鲍秉坤;邓海松;李海波;: "面向智能驾驶视觉感知的对抗样本攻击与防御方法综述", 南京信息工程大学学报(自然科学版), no. 06, 28 November 2019 (2019-11-28) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cao et al. | A many-objective optimization model of industrial internet of things based on private blockchain | |
| CN113762530B (en) | Precision feedback federal learning method for privacy protection | |
| CN108574668B (en) | DDoS attack flow peak value prediction method based on machine learning | |
| CN111625820A (en) | Federal defense method based on AIoT-oriented security | |
| CN113313264B (en) | Efficient federal learning method in Internet of vehicles scene | |
| CN113422695B (en) | Optimization method for improving robustness of topological structure of Internet of things | |
| CN114363043B (en) | Asynchronous federal learning method based on verifiable aggregation and differential privacy in peer-to-peer network | |
| CN114257428B (en) | Encryption network traffic identification and classification method based on deep learning | |
| CN110381509B (en) | Combined authentication method and server suitable for dynamic connection scene | |
| CN112767226A (en) | Image steganography method and system based on GAN network structure automatic learning distortion | |
| CN114448660B (en) | Internet of things data access method | |
| CN113486337B (en) | Network security situation element identification system and method based on particle swarm optimization | |
| CN107016212A (en) | Intention analysis method based on dynamic Bayesian network | |
| CN107317756A (en) | A kind of optimal attack paths planning method learnt based on Q | |
| CN115225383A (en) | Online Web site defense method based on countermeasure patch | |
| CN110784487B (en) | SDN node defense method based on data packet sampling inspection model | |
| CN106603294A (en) | Comprehensive vulnerability assessment method based on power communication network structure and state | |
| CN113726809B (en) | Internet of things equipment identification method based on flow data | |
| CN115021986A (en) | Construction method and device for Internet of things equipment identification deployable model | |
| CN113938309A (en) | DoS attack traffic generation system and method based on DCGAN network | |
| CN115361221B (en) | Website fingerprint identification defense system and method based on data poisoning | |
| CN114528596A (en) | Method for enhancing expandability of directed acyclic graph block chain based on deep learning | |
| CN118432893A (en) | Multi-sample website fingerprint defense method based on challenge sample filling | |
| CN110020087A (en) | A kind of distributed PageRank accelerated method based on similarity estimation | |
| CN116401071B (en) | Resource allocation method and system for edge calculation |
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 |