Table 3 An overview of trust strategies for a recommendation in the Physical layer of IoT
From: Trust-based recommendation systems in Internet of Things: a systematic literature review
Reference | Advantages | Disadvantages | Evaluation technique | Evaluation environment |
|---|---|---|---|---|
Pinto et al. [53] | Assure security and almost intact real-time properties Preempt Linux execution, even performing an IRQ | Applicability for resource-constrained edge devices not implemented Not integrated with hardware trust anchors to tighten security | Â | On a ZedBoard targeting a dual ARM Cortex-A9 running at 600Â MHz |
Cao et al. [54] | Improve transparent and traceable data usage Ensure owners’ obligations and constraints on data usage Provide data access as well as decision explanation Deal with rule conflict Policy composition | Lack of a concrete solution for trustworthy data sharing Not responding query on real-time Not involve end-users in evaluation to ensure usability Not share data on open standard APIs | DPWS and CoAP simulator | SPINdle-based jDUPO Visualiztion tool prototype |
Nieto and Lopez [56] | Multi-hop communications reduce collisions risk while saving energy Convergence, reduce false positives, scalable QoS mechanisms and used in resource-constrained network | Protocol stack has an interface for technologies in MN Sensor’s power consumption to connect to MANET interoperability problems, the complex path towards cooperation |  | g MATLAB and DOT files |
Shirvanimoghaddam et al. [57] | Non-orthogonal multiple access techniques handle a huge number of Cellular devices 3GPP wide area solution on cellular low-power NOMA better throughput RA stage removal and same channel transition Radio resources utilization, and remove signaling overhead | No practical implementation challenges massive IoT in cellular networks:  Device cost  Battery life  Coverage  Scalability  Diversity Challenges massive NOMA in cellular IoT:
  Traffic  Power  Code design  SIC  User fairness | Not mentioned |  |
Zafar et al. [59] | Identifies secure provenance for trust | Runtime overhead in dynamic instrumentation | No evaluation | Â |
Litescu et al. [60] | Twenty percent of participants as optimal data sources Lower information precision to participants noise improves traffic situation with massive participants | Not implemented under realistic traffic network and human behavior Performance decreases with huge involvement as sources/consumers | Poisson process | Â |
Ali et al. [61] | Central trust management lowered the sensors’ energy cost, computation and storage overhead Variable forgetting factor protect against on–off attack High delivery ratios and good data aggregation at base station | Difficulty in contrasting this scheme’s performance with others Trust is centrally handled in base station |  | Using C-Language |
Tan et al. [17] | Effective secure routes Reasonable packet delivery rate, latency, and overhead | To hinder evil nodes, FPNT-OLSR creates longer paths and bigger average latency FPNT-OLSR(R) overhead is more than OLSR’s | OLSR simulation | MoSim MATLAB |
Kang et al. [34] | Protect user privacy and security in IAM | The linear empirical threshold for a tainted hit to IAM App graphics not convenient | TaintDroid | IAM prototype |
Sicari et al. [32] | Secure mobile devices Communication integration in middleware security Collaborative interaction of smart devices | The lack of computing resources and ad hoc nature Indeterminate IoT taxonomy | No simulation or implementation | Â |
Ali et al. [31] | Concurrently measures static and dynamic behaviors Detects deviations Supports multiple stakeholders’ privacy Decreases log size and reduce network overhead Improves detection accuracy | Log size increases with window size | Trusted platform module (TPM) LSM in Linux Kernel Raspberry Pi | Machine learning WEKA |
Køien [67] | Modeling human-to-device trust | TNA-SL was not feasible and practical Cannot cost-effectively mimic actual scenarios | No simulation |  |
Asthana et al. [69] | Consider budget and resource capacity Constraints When no tools exist, give feedback to technological developers | Limited samples Not evaluated with real data and condition such as an individual’s mobility, background, personal preferences, financial factors |  | Evaluated in Weka library |