In this presentation we introduce a new decoding algorithm for general linear block codes and apply it to cryptanalyze McEliece public-key and related systems. The proposed algorithm is of a statistical nature and exploits the statistical information provided by the classical syndrome decoding of linear block codes. The decryption process is performed in two stages: a pre-computation and storage of a number of vectors from the dual code with certain structure and a statistics performed on these vector. These statistics are shown to provide significant information about the location of the errors. Results of calculations show that the McEliece system with its original length at any code rate is not secure.

---