Research Article
Malicious Code Variant Identification Based on Multiscale Feature Fusion CNNs
Table 3
Comparative summary of MFFC algorithm with previous malware classification algorithms.
| Method | Year | Technique | Accuracy (%) | Precision (%) | Recall (%) | F1 score (%) |
| Nataraj et al. [6] | 2011 | ML | 97.18 | — | — | — | SPAM-GIST [10] | 2016 | ML | 97.40 | — | — | — | DL + SVM [17] | 2017 | DL + ML | 84.92 | — | — | — | Vgg-verydeep-19 [16] | 2017 | DL | 97.32 | — | — | — | GIST + SVM [18] | 2018 | ML | 92.20 | 92.50 | 91.40 | — | GIST + KNN [18] | 2018 | ML | 91.90 | 92.10 | 91.70 | — | GLCM + SVM [18] | 2018 | ML | 93.20 | 93.40 | 93.00 | — | GLCM + KNN [18] | 2018 | ML | 92.50 | 92.70 | 92.30 | — | DRBA + CNNs [18] | 2018 | DL | 94.50 | 96.60 | 88.40 | — | LGMP + KNN [11] | 2019 | ML | 98.40 | — | 98.20 | — | NSGA-II + CNNs [19] | 2019 | DL | 97.60 | 97.60 | 88.40 | — | Venkatraman [22] | 2019 | DL | 96.30 | 91.80 | 91.50 | 91.60 | Gibert et al. [28] | 2019 | DL | 98.50 | 98.00 | 98.00 | 98.00 | IMCFN [29] | 2020 | DL | 98.82 | 98.85 | 98.81 | 98.75 | DEAM-DenseNet [30] | 2021 | DL | 98.50 | 96.90 | 96.60 | 96.70 | MFFC | 2021 | DL | 98.72 | 98.86 | 98.72 | 98.73 |
|
|
ML: machine learning; DL: deep learning.
|