Algorithm 2 Distance-based Sampling with Adaptive Threshold Algorithm (DSAT)
Input: D = {Di|1 ≤ i ≤ N, i ∈ Z}, T, C and ε.
Output: D̃ = {D̃i|1 ≤ i ≤ N, i ∈ Z}
1:	Run step 1,2,3,4 in Algorithm 1;
2:	Skip the first M timestamps;
3:	for each time point ti with i > M do
4:	if count ≥ C, then set D̃i = D̃j
5:	Set $\stackrel{\sim }{d}(D_i,{\stackrel{\sim }{D}}_j)=d(D_i,{\stackrel{\sim }{D}}_j)+\mathit{\text{Lap}}\left(\frac{2C\mathrm{\Delta }}{{\epsilon }_1}\right)$
6:	Set $E_i=|\frac{\mathit{\text{count}}}{i}-\frac{C}{N}|,e_i=\frac{|E_i-\delta |}{\delta }$, and ui = θei;
7:	if $\frac{\mathit{\text{count}}}{t}-\frac{C}{N}\le 0$, then set T̃i = max{0, T̃i−1 − ui}
8:	else set T̃i = min{2, T̃i−1 + ui};
9:	if d̃(Di, D̃j) ≥ T̃i then
10:	release a DP dataset D̃i at ti with $\frac{{\epsilon }_2}{C}$ budget, and set count = count + 1, j = i;
11:	else
12:	release D̃j;
13:	end if
14:	if i == N and count < C then
15:	release D̃N with all remaining privacy budget;
16:	end if
17:	end for