Fig 1.
The classic TSR models in S-SNs.
A: Reputation-based. B: Local trust-based.
Table 1.
Intuitive comparisons between our IRLT model and other TSR models in S-SNs.
Fig 2.
Our IRLT model for TSR in S-SNs.
①: TSR request. ② and ③: Candidate services. ④ and ⑤: Candidate services with local trust values. ⑥: Candidate services with reputation values. ⑦ and ⑧: Top-k recommendation list.
Table 2.
A simple example of our multi-QoS based filtering method.
Table 3.
A simple example of aggregation trust evaluation.
Table 4.
The values of parameters in experiments.
Fig 3.
The variations of three kinds of DCG metric values with k in honest environment.
A: DCG-LT. B: DCG-RP. C: DCG-FT.
Fig 4.
The variation of total DCG-FT metric value of malicious services with k in malicious environment with ballot stuffing attack.
A: PCSC = 0%. B: PCSC = 20%. C: PCSC = 40%. D: PCSC = 60%. E: PCSC = 80%. F: PCSC = 100%.
Fig 5.
The variation of total DCG-FT metric value of honest services with k in malicious environment with bad mouthing attack.
A: PCSC = 0%. B: PCSC = 20%. C: PCSC = 40%. D: PCSC = 60%. E: PCSC = 80%. F: PCSC = 100%.
Fig 6.
The variations of total DCG-FT metric values of honest services (HS) and malicious services (MS) with k in malicious environment with ballot stuffing and bad mouthing attacks.
A: PCSC = 0%. B: PCSC = 25%. C: PCSC = 50%. D: PCSC = 75%. E: PCSC = 100%.