1. TRUST-BASED DECISION MAKING FOR HEALTH IoT SYSTEM Hamid Al-Hamadi and Ing-Ray Chen present by Ning Wang

2. Introduction of Health IoT System Model Protocol Design
OUTLINE
Introduction of Health IoT
System Model
Protocol Design
Performance Evaluation
Summary

3. Introduction of Health IoT1
Introduction of Health IoT
IoT, HIoT, Environmental H IoT

4. IoT and Health IoT IoT Health IoT
is the network of physical devices, vehicles, home appliances, and other items embedded with electronics, software, sensors, actuators and connectivity which enables these things to connect, collect and exchange data.
Health IoT
is a type of IoT used for Healthcare or medical. For example, the remote monitoring of medical parameters, smart hospital services, individual well-being, and emergency site and rescue.

5. Environmental Health IoT Environmental monitor /sensor
IoT and Health IoT
Environmental Health IoT
Environmental monitor /sensor
mobile application in smart phone
Internet/ cloud
The application
environmental parameter measuring
prior knowledge of the environment
assist decision making

6. Motivation
Doctor scarcity
it is difficult for health professionals to personally attend to all patients at all times.
Lack of environmental Knowledge
It is difficult to assess the health risk for patients entering into different physical locations without detailed condition.
Personalization
The current health status should be considered.

7. System model and threat model2
System Model
System model and threat model

8. Decision making model Environmental data Health status data
responsible for maintaining the thresholds data and is what health experts use to interact with the system
A trust management subsystem is responsible for trust and risk calculations and management
It further stores all member sensor readings for future decision making.
A communication subsystem is responsible for handling incoming queries and incoming data.

9. Decision making model
A member intend to enter a location, send a query to CA ask about the safety
Risk calculation base on
health expert and trust management
Give the recommendation

10. Decision making model Z: vulnerability, health classification
Z=1-H, where H is health index
G: Probability of health loss
p: reliability trust of the source
Parameter Z is a member’s health classification by the doctor/medical center. Parameter p is the reliability trust of the source of the sensing data. Parameter G is the possibility of health loss as derived from the sensing data.

11. Decision making model
All the decision points below this graph are considered logically good decisions.

12. The faulty sensors can give incorrect readings due to malfunction.
Threat Model
A malicious member may show untrustworthy behavior to further its interests.
A member may be reluctant to waste its resources for the benefit of the health IoT system.
The faulty sensors can give incorrect readings due to malfunction.
malicious attackers who aim to break down the health IoT system.
No collution
The trust management protocol is to overcome these incorrect readings
and provide the most trustworthy data
thresholds data and is what health experts use to interact with the system
A trust management subsystem is responsible for trust and risk calculations and management
It further stores all member sensor readings for future decision making.
A communication subsystem is responsible for handling incoming queries and incoming data.

13. Rating, query, trust score3
Protocol Design
Rating, query, trust score

14. IoT and Health IoT
Location Rating
Query processing by CA

15. Protocol Design Reply of the query Aggregated rating Aggregated Trust
Learning experience:
It derives the aggregate rating for the location, taking into consideration of the associated trustworthiness scores of raters for the aggregate rating.

16. Mapping the rating data to (p,G) for decision making
Protocol Design
Mapping the rating data to (p,G) for decision making
Alternatively: a list of information for local decision making
(Z,p,G)
A decision: Yes or No
G: Probability of health loss
p: reliability trust of the source
Member 𝑖 can then process this information locally for decision making. An advantage of this method is to enable the user to personalize the receiving information by further applying its own measurements of trust scores towards other entities such as its relatives and close friends which could be given higher trust by default and their information could be considered more trustworthy.

17. Trust Score Computation
Protocol Design
Trust Score Computation
location rating score
rater trust score
witness trust score
trust score RR_k,I
: Location rating trust score
:Rater trust score
: Witness trust score

18. Location rating trust score
Protocol Design
Location rating trust score
Feedback factor
Time relevant
Similarity
Equals to
This keeps the comparison more relevant at the time of assessment (to minimize weight of old data)

19. Protocol Design Rater trust score
compare node k’s feedback with the majority of feedbacks
This keeps the comparison more relevant at the time of assessment (to minimize weight of old data)

20. Protocol Design Witness trust score
the CA examines the trustworthiness of its received ratings by all members
Two benefits:
1. Members that have been seen by other members in the reported location gain trust.
2. It can detect misbehavior in reported location ratings,
It relies on data from members that vouch for the correctness of the claim that another member was in fact in a location at a specific time.

21. Protocol Design Member: query and feedback CA: Trust management
Receive a query,
Calculate the aggregated rating and trust(G, p)
Calculate Z with (G,p), then get (Z,G,p).
reply Yes or No based on Decision plane.
Query the HIoT system whether to enter a location
Enter the location with approval, give feedback and witness list.
Three trust scores, update the overall trust score.
Compare the trust score with a threshold: evict or not
It relies on data from members that vouch for the correctness of the claim that another member was in fact in a location at a specific time.
Member: query and feedback
CA: Trust management
CA: reply a query

22. Performance Evaluation4
Performance Evaluation

23. Performance Evaluation
The parameters for performance evalution
we perform ns3 simulation for performance evaluation of our trust-based decision making protocol, and conduct a comparative analysis with two baseline decision making protocols.

24. Performance Evaluation
converges fairly quickly as more information is collected as time progresses. The convergence time increases as 𝑃 𝑚 increases. However, we see that CDR eventually converges to a high value even when 𝑃 𝑚 is as high as 30%.
Figure 10 demonstrates the effectiveness of our strategy in identifying and evicting untrustworthy users.
we show how our trust system measures the trust score of a good node, 𝑘, turning into malicious after 𝑇 𝑐𝑜𝑚𝑝 (ranging from 5hrs to 15hrs) is elapsed.
Figure 12 shows the effect of a member’s health status on CDR.
An interesting trend is that the lower the health of the member (lower H) the more sensitive it is to attacks by malicious nodes (higher 𝑃 𝑚 ) and

25. Performance Evaluation
The protocols
No Trust (NT): it does not have trust management.
No Member Health (NMH): it does not relate 𝑍 with 𝐺.
converges fairly quickly as more information is collected as time progresses. The convergence time increases as 𝑃 𝑚 increases. However, we see that CDR eventually converges to a high value even when 𝑃 𝑚 is as high as 30%.
Figure 10 demonstrates the effectiveness of our strategy in identifying and evicting untrustworthy users.
we show how our trust system measures the trust score of a good node, 𝑘, turning into malicious after 𝑇 𝑐𝑜𝑚𝑝 (ranging from 5hrs to 15hrs) is elapsed.
Figure 12 shows the effect of a member’s health status on CDR.
An interesting trend is that the lower the health of the member (lower H) the more sensitive it is to attacks by malicious nodes (higher 𝑃 𝑚 ) and

26. 5
Summary

27. Summary Conclusion Future extension
This paper proposed trustbased health IoT protocol considers risk classification, reliability trust, and loss of health probability as three design.
This paper develops a trust computation protocol for a health IoT system to assess the reliability trust of individual IoT devices.
Future extension
Change the centralized structure to a distributed structure.
Consider the collusion of attackers.

28. Thanks!