The properties of cyclostationary and periodically correlated models of noisy signals have been analyzed. The fundamental principles of the theory of linear periodic random processes have been studied, along with the probabilistic characteristics of their periodicity. Solutions to the problem of decomposing cyclostationary noisy signals into Fourier series have been provided. The possibilities of identifying noisy signals using characteristics of periodic autoregression, moving sum, and periodic autoregressive moving sum models have been examined. The necessary conditions that the kernel and generating process with independent increments in the representation of a conditional linear random process must satisfy in order for it to be periodically correlated or cyclostationary have been justified. The method of conditional characteristic functions has been used for justification. Additionally, the properties and characteristics of identifying the spectral model of noisy signals in the form of a harmonized periodic random process have been discussed.