Research Article
[Retracted] Privacy Protection Technology Based on Machine Learning and Intelligent Data Recognition
Table 3
Comparison of the classification accuracy and attack accuracy between our method and CNN under different parameters in MNIST dataset.
| Data size | Epoch | Model accuracy | Attack success rate | Model accuracy (CNN) | Attack success Rate (CNN) (%) |
| 10000 | 10 | 98.57 (%) | 11.76 (%) | 95.68 | 98.05 (%) | 20000 | 10 | 98.42 | 12.26 | 96.12 | 88.76 | 30000 | 10 | 97.81 | 12.43 | 94.75 | 67.52 | 10000 | 25 | 98.54 | 13.05 | 95.04 | 79.86 | 20000 | 25 | 99.07 | 10.94 | 96.08 | 65.42 | 30000 | 25 | 97.88 | 12.72 | 94.87 | 60.17 | 10000 | 50 | 98.35 | 13.28 | 95.59 | 75.84 | 20000 | 50 | 98.12 | 11.74 | 94.25 | 62.56 | 30000 | 50 | 99.06 | 12.82 | 96.06 | 53.93 |
|
|