We report the discovery of a massive and dense transiting planet CoRoT-27b on a 3.58 day orbit around a 4.2 Gyr-old G2 star. The planet candidate was identified from the CoRoT photometry, and was confirmed as a planet with ground-based spectroscopy. The confirmation of the planet candidate is based on radial velocity observations combined with imaging to rule out blends. The characterisation of the planet and its host star is carried out using a Bayesian approach where all the data (CoRoT photometry, radial velocities, and spectroscopic characterisation of the star) are used jointly. The Bayesian analysis includes a study whether the assumption of white normally distributed noise holds for the CoRoT photometry, and whether the use of a non-normal noise distribution offers advantages in parameter estimation and model selection. CoRoT-27b has a mass of $10.39 \pm 0.55$ $\mathrm{M}_{\rm Jup}$, a radius of $1.01 \pm 0.04$ $\mathrm{R}_{\rm Jup}$, a mean density of $12.6_{-1.67}^{+1.92}$ $\mathrm{g\,cm^{-3}}$, and an effective temperature of $1500 \pm 130$ K. The planet orbits around its host star, a 4.2 Gyr-old G2-star with a mass $M_{\star}=1.06$ $M_{\odot}$, and a radius $R_{\star}=1.05$ $R_{\odot}$, on a $0.048 \pm 0.007$ AU orbit every 3.58 days. The radial velocity observations allow us to exclude highly eccentric orbits, namely, $e<0.065$ with a 99% confidence. Given its high mass and density, theoretical modelling of CoRoT-27b is demanding. We identify two solutions with heavy element mass fractions of $0.11\pm0.08$ $\mathrm{M_{\oplus}}$ and $0.07\pm0.06$ $\mathrm{M_{\oplus}}$, but even solutions void of heavy elements cannot be excluded. We carry out a secondary eclipse search from the CoRoT photometry using a method based on Bayesian model selection, but conclude that the noise level is too high to detect eclipses shallower than 9% of the transit depth.