This study first proposes a concept of generalized stability (GST) for discrete chaos system, which is the generalization of chaos generalized synchronization (CGS). Then this study sets up a constructive theorem of GST for discrete chaos system, which provides a general representation of GST in discrete chaos system. Using the theorem designs an 8-dimensional GST system consisting of a driving chaotic system and a driven chaotic system. Numerical simulation verifies the chaotic dynamic behaviors of such GST system, which is used to design a chaotic pseudorandom number generator (CPRNG). Using FIPS 140-2 test suite and G FIPS 140-2 test suite test the randomness of four 1,000-key streams consisting of 20,000 bits generated respectively by the CPRNG, the RC4 algorithm and the ZUC algorithm. The results show that the randomness performances of the CPRNG is promising, and suggest that the statistical properties of the randomness of the sequences generated via the CPRNG and the two algorithms do not have significant differences. As an application, using the sequences generated via the CPRNG and a stream encryption scheme with avalanche effect (SESAE) encrypts an RGB image. The results show that the encrypted RGB image have significant avalanche effects, and suggest the CPRNG is a qualified candidate for the stream encryption scheme with avalanche effect.