Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode
The threat model of deep-learning-based data theft in data sandbox model was analyzed in detail, and the degree of damage and distinguishing characteristics of this attack were quantitatively evaluated both in the data processing stage and the model training stage.Aiming at the attack in the data pr...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Department of Journal on Communications
2021-11-01
|
| Series: | Tongxin xuebao |
| Subjects: | |
| Online Access: | http://www.joconline.com.cn/zh/article/doi/10.11959/j.issn.1000-436x.2021215/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841539219258867712 |
|---|---|
| author | Hezhong PAN Peiyi HAN Xiayu XIANG Shaoming DUAN Rongfei ZHUANG Chuanyi LIU |
| author_facet | Hezhong PAN Peiyi HAN Xiayu XIANG Shaoming DUAN Rongfei ZHUANG Chuanyi LIU |
| author_sort | Hezhong PAN |
| collection | DOAJ |
| description | The threat model of deep-learning-based data theft in data sandbox model was analyzed in detail, and the degree of damage and distinguishing characteristics of this attack were quantitatively evaluated both in the data processing stage and the model training stage.Aiming at the attack in the data processing stage, a data leakage prevention method based on model pruning was proposed to reduce the amount of data leakage while ensuring the availability of the original model.Aiming at the attack in model training stage, an attack detection method based on model parameter analysis was proposed to intercept malicious models and prevent data leakage.These two methods do not need to modify or encrypt data, and do not need to manually analyze the training code of deep learning model, so they can be better applied to data theft defense in data sandbox mode.Experimental evaluation shows that the defense method based on model pruning can reduce 73% of data leakage, and the detection method based on model parameter analysis can effectively identify more than 95% of attacks. |
| format | Article |
| id | doaj-art-9e262ecf2510466abdc0aa78f742f444 |
| institution | Kabale University |
| issn | 1000-436X |
| language | zho |
| publishDate | 2021-11-01 |
| publisher | Editorial Department of Journal on Communications |
| record_format | Article |
| series | Tongxin xuebao |
| spelling | doaj-art-9e262ecf2510466abdc0aa78f742f4442025-01-14T07:23:08ZzhoEditorial Department of Journal on CommunicationsTongxin xuebao1000-436X2021-11-014213314459746072Threat analysis and defense methods of deep-learning-based data theft in data sandbox modeHezhong PANPeiyi HANXiayu XIANGShaoming DUANRongfei ZHUANGChuanyi LIUThe threat model of deep-learning-based data theft in data sandbox model was analyzed in detail, and the degree of damage and distinguishing characteristics of this attack were quantitatively evaluated both in the data processing stage and the model training stage.Aiming at the attack in the data processing stage, a data leakage prevention method based on model pruning was proposed to reduce the amount of data leakage while ensuring the availability of the original model.Aiming at the attack in model training stage, an attack detection method based on model parameter analysis was proposed to intercept malicious models and prevent data leakage.These two methods do not need to modify or encrypt data, and do not need to manually analyze the training code of deep learning model, so they can be better applied to data theft defense in data sandbox mode.Experimental evaluation shows that the defense method based on model pruning can reduce 73% of data leakage, and the detection method based on model parameter analysis can effectively identify more than 95% of attacks.http://www.joconline.com.cn/zh/article/doi/10.11959/j.issn.1000-436x.2021215/data sandboxdata theftsecurity of AI |
| spellingShingle | Hezhong PAN Peiyi HAN Xiayu XIANG Shaoming DUAN Rongfei ZHUANG Chuanyi LIU Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode Tongxin xuebao data sandbox data theft security of AI |
| title | Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode |
| title_full | Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode |
| title_fullStr | Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode |
| title_full_unstemmed | Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode |
| title_short | Threat analysis and defense methods of deep-learning-based data theft in data sandbox mode |
| title_sort | threat analysis and defense methods of deep learning based data theft in data sandbox mode |
| topic | data sandbox data theft security of AI |
| url | http://www.joconline.com.cn/zh/article/doi/10.11959/j.issn.1000-436x.2021215/ |
| work_keys_str_mv | AT hezhongpan threatanalysisanddefensemethodsofdeeplearningbaseddatatheftindatasandboxmode AT peiyihan threatanalysisanddefensemethodsofdeeplearningbaseddatatheftindatasandboxmode AT xiayuxiang threatanalysisanddefensemethodsofdeeplearningbaseddatatheftindatasandboxmode AT shaomingduan threatanalysisanddefensemethodsofdeeplearningbaseddatatheftindatasandboxmode AT rongfeizhuang threatanalysisanddefensemethodsofdeeplearningbaseddatatheftindatasandboxmode AT chuanyiliu threatanalysisanddefensemethodsofdeeplearningbaseddatatheftindatasandboxmode |