Using the latest AMS-02 cosmic ray antiproton flux data, we search for potential dark matter annihilation signal. The background parameters about the propagation, source injection, and solar modulation are not assumed a priori, but based on the results inferred from the recent B/C ratio and proton data measurements instead. The possible dark matter signal is incorporated into the model self-consistently under a Bayesian framework. Compared with the astrophysical background only hypothesis, we find that a dark matter signal is favored. The rest mass of the dark matter particles is ∼ 20 − 80 GeV and the velocity-averaged hadronic annihilation cross section is about (0.2 − 5) × 10 −26 cm 3 s −1 , in agreement with that needed to account for the Galactic center GeV excess and/or the weak GeV emission from dwarf spheroidal galaxies Reticulum 2 and Tucana III. Tight constraints on the dark matter annihilation models are also set in a wide mass region. PACS numbers: 95.35.+d,Introduction -The precise measurements of cosmic ray (CR) anti-particle spectra by space-borne instruments, such as PAMELA and AMS-02, provide very good sensitivity to probe the particle dark matter (DM) annihilation or decay in the Milky Way. The CR antiprotons, primarily come from the inelastic collisions between the CR protons (and Helium) and the interstellar medium (ISM), are effective to constrain the DM models [1][2][3]. Recent observations of the antiproton fluxes [4][5][6] are largely consistent with the expectation from the CR propagation model, leaving very limited room for the annihilation or decay of DM [2,[7][8][9][10].There are several sources of uncertainties in using antiprotons to constrain DM models. The largest uncertainty may come from the propagation parameters. Usually the secondary-to-primary ratio of CR nuclei, such as the Boronto-Carbon ratio (B/C), and the radioactive-to-stable isotope ratio of secondary nuclei, such as the Beryllium isotope ratio 10 Be/ 9 Be, are used to determine the propagation parameters [11,12]. Limited by the data quality, the constraints on the propagation parameters are loose [13,14]. Even the effect on the background antiproton flux due to uncertainties of propagation parameters is moderate, the flux from the DM component depends sensitively on propagation parameters [15]. Additional uncertainties include the injection spectrum of the CR nuclei, solar modulation, and hadronic interaction models [8]. Those uncertainties make the DM searches with antiprotons inconclusive [16,17].Given the new measurements of the proton, Helium, and B/C data by , improved constraints on the propagation and source injection parameters can be obtained through global Bayesian approaches [22][23][24][25]. With these data, we conduct a global study to determine the propagation, injection, and solar modulation parameters si- * The corresponding author: yuanq@pmo.ac.cn † The corresponding author: yzfan@pmo.ac.cn multaneously using the Markov Chain Monte Carlo (MCMC) method [26]. These "background" parameters and their like...