Algorithm 1: Fusion-based trust anomaly detection algorithm.

Input:${\mathsf{\zeta }}_i$, $\mathsf{\rho }$, ${\mathsf{\mu }}_{*}$ Output:${\mathsf{\eta }}_t^i$, ${\mathsf{\varrho }}_S^R$, ${\mathsf{\varsigma }}_t$, $S^2$ begin 1:    detection system is ready 2:    return${\mathsf{\eta }}_t^i$ 3:    for n = 1: length(${\mathsf{\zeta }}_i$) 4:            $T_n={\mathsf{\zeta }}_i\left(n\right), T_{n+1}={\mathsf{\zeta }}_i(n+1)$ 5:            $∆T=T_n-T_{n+1}$ 6:            if $∆T<\mathsf{\rho }$ then 7:                be considered normal 8:            else 9:                $T_{n+1}$ is added to ${\mathsf{\xi }}_t\left(n\right)$ 10:      end for 11:      foreach T in ${\mathsf{\xi }}_t\left(n\right)$ 12:         if $\frac{\mathsf{\partial }{T}_n}{\mathsf{\partial }a}-\frac{\mathsf{\partial }∆T}{\mathsf{\partial }a}=0, \frac{\mathsf{\partial }{T}_n}{\mathsf{\partial }b}-\frac{\mathsf{\partial }∆T}{\mathsf{\partial }b}\ne 0$ 13:               anomaly source is in communication layer 14:         elseif $\frac{\mathsf{\partial }{T}_n}{\mathsf{\partial }a}-\frac{\mathsf{\partial }∆T}{\mathsf{\partial }a}\ne 0, \frac{\mathsf{\partial }{T}_n}{\mathsf{\partial }b}-\frac{\mathsf{\partial }∆T}{\mathsf{\partial }b}=0$ 15:               anomaly source is in transaction layer 16:         elseif $\frac{\mathsf{\partial }{T}_n}{\mathsf{\partial }a}-\frac{\mathsf{\partial }∆T}{\mathsf{\partial }a}=0, \frac{\mathsf{\partial }{T}_n}{\mathsf{\partial }b}-\frac{\mathsf{\partial }∆T}{\mathsf{\partial }b}=0$ 17:               anomaly source is in both layers 18:         return ${\mathsf{\varrho }}_S^R$ 19:      end for 20:      $S^2=\frac{1}{n-1}{\sum }_{i=1}^n{({NT}_i-{\mathsf{\mu }}_{*})}^2$ 21:      return $S^2$, ${\mathsf{\varsigma }}_t$ end