Review Article
Deep Learning Methods for Malware and Intrusion Detection: A Systematic Literature Review
Table 11
Analysis of the sustainability of android-based malware detection techniques.
| Ref. | Evolvability/ability of identifying new malware? (yes/no) | DL model need updating/retraining? (yes/no) | Sustainability/resilience against evolution? (yes/no) | Sustainable up to years/accuracy | Initial accuracy/F1 score |
| Towards sustainable android malware detection [132] | Yes | After 5 years | Yes | 5 (82%) | Above 93% | MaMaDroid [133] | Yes | After 2 years | Yes | 2 (87%) | 99% | Frequency analysis model (FAM)—a variant of MaMaDroid [134] | Yes | After several years | Yes | Several (76%) | 81% | DroidSpan [135] | Yes | After 7 years | Yes | 1–7 (21–37% superior than competitor) | 6–32% superior than competitor | RevealDroid [136] | Yes | After 3 years | Yes | 3 (87%) | 98% | DroidCat [137] | Yes | After 9 years | Yes | 9 (97%) | 97% | G-Droid [138] | Yes | — | Yes | — | 98.99% |
|
|