From: Trust-based recommendation systems in Internet of Things: a systematic literature review
Reference | Advantages | Disadvantages | Evaluation technique | Evaluation environment |
---|---|---|---|---|
Wang et al. [74] | Overcomes the DoS threats by selfish nodes Avoids DDoS threats by crowdsourcing | Non-cooperative parties reduce efficiency Non-malicious members received Lower may wrong results provide by honest participants | No simulation | Â |
Fortino et al. [75] | Mutual benefit of single agent engagement in a group By group formation, the untrusted agents, gradually substituted with trusted ones Local reputation avoids global mechanism’s overhead Good performance achieved with agents’ small portion of involvement | Few statistical parameters were concerned With a minimum number of examined group, a few untrusted agents exit which lessened not totally removed with rising numbers | Poisson distribution | Not mentioned |
Garcia-de-Prado et al. [78] | Facilitates communications and data delivery between C-IoT layers Avoids edge nodes waste resource Save expense in cloud real-time data processing | Not process many events per second in fog nodes Deter user profile and experience | Esper CEP | Mule open source ESB MQTT Mosquitto broker—Eclipse Case study: ir4HealthAdmin |
Sfar et al. [81] | Adapted to any real environment to Improve productivity intensify issues of security and privacy | Theoretical rigor limitation Do not consider: auto-immunity, safety, reliability, and responsibility | Â | Case study |
Ouaddah et al. [82] | OM-AM access control solutions Considers centralized and decentralized approaches | Usability aspects are not extended No implementation on privacy-preserving access control framework | No simulation | Â |
Roman et al. [83] | Cooperation diverse IoT entities despite lack of central systems Probability to pinpoint the problem origins Implements privacy and scalability Push/pulls data when needed | Key management between the limited device Overhead caused by incoming connections Concerns about internet protocols adaptability with context | No simulation | Â |
Mahalle et al. [84] | Guarantees scalability, flexibility, and energy efficiency Devices proliferation does not degrade performance Avoids communication in low trust nodes More power conservation and high residual source | A mathematical model not implemented in real time RFID Not organized with capability-based access control | Mamdani logic | MATLAB 7.0. 
 NS2 |
Sicari et al. [86] | Validated by real-time open data feeds NOS detects:  Data confidentiality, source privacy, and integrity violations;  Unauthorized access  Robustness of key management  Replay or routing attacks Without restarting the whole system add new modules, duplicate or remove available ones Independent of the data model and application domain | Low reliability regarding confidentiality, integrity, and privacy, weak accuracy and precision |  | Open-source Mosquitto Visualization service not mentioned |
Hellaoui et al. [87] | Energy conservation and yet secure Vast participation by tracking entities’ attributes Deterred adversaries to alternate between forwarding a (un)authenticated messages | Not consider untrustworthy recommendations Not evaluated in lossy networks | HMAC MD5 | Cooja, the Contiki OS simulator |
Azad et al. [88] | Assess nodes’ reputation: (1) Independent of third-person and stay away of:   Single point of failure   Be in middle of threat   Privacy destruction (2) Use aggregated behavior-based weights (3) Encryption/decryption promise privacy preservation (4) Avoid fraudulent response by NIZK proof Security is ensured by Decisional Diffie Hellman Trust value is concealed and could not abused to derive relations Bandwidth and storage capacity of RSU is sensible Free infrastructure of physical layer and performed in session layer | Delay occurs for database exploration and caused by wireless connection among objects | NIZK proof CPU 3.6 GHz core i7, 8 GB memory | Java program |
Yu et al. [90] | Lowers the energy usage due to trust exchange among adjacent members Ensures accuracy and minimizes recommended trust Obtains more robustness and adaptability Security of packets forwarding and defeating attacks | Assumptions: Densely deployed sensors are prohibited to move freely Communicate under maximum power Attacker intercept any nodes has communication capabilities as normal nodes The communication channel is symmetric | Â | Simulation in MATLAB |
Khan et al. [93] | Explored trust management on SIoT Define SCIoT Propose an attack model Classifies probable threats Highlighted relevant challenges | The proposed issue remains unresolved | Not implemented | Â |
Ing-Ray Chen [94] | Dynamic trust management Accurate trust assessment or minimizing trust bias Maximize application performance totally distributed without the need for centralized entities | Not consider the thing-to-thing autonomous Social relation Disregard the caching usage to relieve constraint storage After status changes recommendations do not improve trust convergence | Â | ns3 |
Dwarakanath et al. [95] | In decentralized sources, privacy-aware collaboration consumes less than 0.5% battery for execution | On–off adversaries lengthen delay in gaining trust values Combats collusion and on–off threats of a mischievous minority Not practicable for single device data |  | Google Nexus 5 smartphones CEPsim |
Wang et al. [97] | Screens dishonest nodes Desirable convergence, accuracy, and resiliency Compared with P2P gives more accurate trust prediction and resists against collusion Resilient to ballot-stuffing, bad-mouthing Applicability in a hostile and noisy environments | The overhead for SOANET with constraint memory capacity Trust prediction takes minutes rather than seconds for less powerful node | Random Waypoint mobility (RWM) | MATLAB |