Exploration of Impactful Countermeasures on IoT Attacks

IoT threat mitigation engine empowered by artificial intelligence multi-objective optimization

“…The third class includes approaches where the selection of mitigation actions is based on the optimization of the value a single KPI or transforming the problem to a single objective (SO) problem e.g. (Chehida et al, 2020) where the authors select countermeasures based on minimizing the cost to deploy them.…”

“…In the context of IoT networks specifically, KPI optimization for mitigation action selection has become a recent topic of interest. Three solutions involving SO approaches for mitigation selection have been identified, either in generic IoT networks (Chehida et al, 2020;Garzia et al, 2017) or smart energy grid systems (Zonouz and Haghani, 2013). Two IoT specific studies using MO have been identified i.e.…”

“…Literature review results on the selection of mitigation actions for cyber-attacks Approach Category Approaches No automation N=3, Visualization (Gonzalez- Granadillo et al, 2015;Granadillo et al, 2012;Gonzalez-Granadillo et al, 2017) Heuristic Approach N=3, (Chung et al, 2013;Kotenko and Doynikova, 2016;Doynikova and Kotenko, 2018) Single-objective Optimization N=8, Ant Colony (Wang et al, 2013), Bellman Method (Zonouz et al, 2009;Zonouz and Haghani, 2013;Miehling et al, 2015), Genetic Algorithm (Poolsappasit et al, 2012;Garzia et al, 2017;Li et al, 2018;Chehida et al, 2020) Multi-objective Optimization N=10, Branch & Bound Integer Optimization (Roy et al, 2012),Genetic Algorithm , NSGA-II algorithm (Dewri et al, 2012;Rachedi and Benslimane, 2016;Lee et al, 2017;Hasan et al, 2018;Enoch et al, 2019), Single Additive Weighting & Weighted Product method (Shameli-Sendi and Dagenais, 2015;Shameli-Sendi et al, 2018), Tabu Search algorithm (Viduto et al, 2012) of the objective functions examined can be improved without the value of some other objective deteriorating. Solutions with this property are called non-dominated or Pareto optimal.…”

“…The authors show that a technique called Integer programming can compete with a Simple GA in the runtime required to find the optimal solution Set. (Chehida et al, 2020) The authors present a list of 11 attacks and 10 countermeasures along with their cost and other security related parameters and then try to find a optimal policy balanced between cost and risk in a finite time duration.…”

Time Series Network Data Enabling Distributed Intelligence—A Holistic IoT Security Platform Solution

“…Such an example is presented in [30]: based on predefined scenarios and values, the system chooses sets of predefined responses. The third group includes approaches where the selection of mitigation actions is based on the optimization of the value of a single KPI or transforming the problem to a single objective problem, e.g., [31], where the authors select attack responses based on minimizing the cost of deploying them. Finally, the fourth class includes the approaches where the selection of mitigation actions is based on the optimization of the values of multiple KPIs which might be antagonistic to each other but better describe the impact of the action on the system.…”