1: \begin{abstract}

2:   Under Markovian assumptions, we leverage a Central Limit Theorem (CLT)

3:   for the empirical measure in the test statistic of the composite

4:   hypothesis Hoeffding test so as to establish weak convergence results

5:   for the test statistic, and, thereby, derive a new estimator for the

6:   threshold needed by the test. We first show the advantages of our

7:   estimator over {an existing estimator} by conducting extensive

8:   numerical experiments. We find that our estimator controls better for

9:   false alarms while maintaining satisfactory detection

10:   probabilities. We then apply the Hoeffding test with our threshold

11:   estimator to detecting anomalies in two distinct applications domains:

12:   one in communication networks and the other in transportation

13:   networks. The former application seeks to enhance cyber security and

14:   the latter aims at building smarter transportation systems in cities.

15: \end{abstract}

16: