|
| Author | Gap analysis for novel approach |
|
| IoT methodologies |
| Zarpelão et al. [5] | The paper identified a gap using ML and IDS based on power consumption. Power consumption metrics will be in the scope of our work. |
|
| Rehman et al. [3] | The paper identified a gap using ML and IDS based on power consumption and hop count for RAOF. Power consumption and hop count will be in the scope of our work. |
|
| Le et al. [4] | The combination of attacks has not been considered. Power consumption and dropped packet features could be used as a novel approach to anomaly-based detection. A combination of IoT attacks along with power consumption and dropped packet will be in the scope of our work. |
|
| MRHOF and OF0 attacks |
| Airehrour et al. [6] | Research failed to detect individual attacks against OF0 and MRHOF. MRHOF and OF0 will be in the scope of our work for each IoT combination attack. |
|
| Airehrour et al. [7] | The paper identified a gap for detecting/isolating a combination of Rank and Sybil Attacks within MRHOF and OF0. A combination of Rank and Sybil Attack along with MRHOF and OF0 will be in the scope of our work for each IoT combination attack. |
|
| Mehta and Parma [8] | The paper identified a gap that possible OF attacks should be detectable. A combination of IoT attacks along with MRHOF and OF0 will be in the scope of our work for each IoT combination attack. |
|
| IDS methodologies and feature selection |
| Sheikhan and Bostani [9] | Research failed to detect unknown attacks using selected features for misuse-based detection. Power consumption and dropped packet metrics will be in the scope of our work. |
|
| Mayzaud et al. [10, 11] | Despite authors claiming their research as a feasible solution for anomaly detection for IoT, there is no evidence of detection for a wide range of attacks beyond DAG. A combination of IoT attacks will be in the scope of our work. |
|
| Lee et al. [12] | OF and MRHOF are not discussed within discussion of detecting malicious activity based on power consumption and network flow. MRHOF and OF0 for each IoT combination attack along with power consumption, hop count, and dropped packet metrics will be in the scope of our work. |
|
| Sousa et al. [13] | The paper discussed OF-FL, CAOF, and other relevant OF routing metrics and then excluded them during simulation. MRHOF and OF0 will be in the scope of our work for each IoT combination attack. |
|
| Napiah et al. [14] | The paper discussed reducing features from 77 to 5 removing power consumption to ensure ML algorithms were efficient. Power consumption metrics along with feature reduction strategy will be in the scope of our work. |
|
| Datasets and ML classifiers |
| Haq [15] | The paper reviews 49 related studies and highlights considerations to be made when developing a ML-IDS. ML approach, classifier methods, suitable algorithms, datasets, and features selection are in the scope of our work. |
|
| Nannan et al. [16] | Research identified a high false alarm rate for anomaly detection. ML approach, classifier methods, suitable algorithms, datasets, and features selection are in the scope of our work. |
|
| Buczak and Guven [17] | KDD 1999 is limited by attacks that have occurred since the dataset was produced. This includes IoT attacks. The identification or development of a novel dataset focused on IoT features and attacks is in the scope of our work. |
|
| Alam et al. [18] | The paper identified little research into the use of conventional ML algorithms with IoT datasets. The identification or development of a novel dataset focused on IoT features and attacks along with the ML approach, classifier methods, suitable algorithms, datasets, and features selection are in the scope of our work. |
|
| Preprocessor and balancing techniques |
| Yin and Gai [19] | The paper reviews 12 datasets and highlights considerations to be made when developing an imbalanced dataset. Preprocessor techniques, feature extension, sampling, as well as train, test, and validate datasets are in the scope of our work. |
|
