Some cryptographical applications use pseudorandom sequences and require that the sequences are secure in the sense that they cannot be recovered by only knowing a small amount of consecutive terms. The security of the sequences is translated into several measurable characteristics. For example they should have a large linear complexity and also a large k-error linear complexity. This thesis focuses on the k-error linear complexity of sequences. Currently, efficient algorithms for computing the k-error linear complexity of a sequence only exist for special classes of sequences, e.g. of period equal to a power of the characteristic of the field. It is therefore useful to find a general and efficient algorithm to compute a good approximation of the k-error linear complexity.