Machine learning-aided and SAT-aided cryptanalysis of symmetric-key primitives
Information security has received more and more attentions in recent decades with the rapid developments of the internet era. Since symmetric cryptographic primitives are widely used in current information systems, doing cryptanalysis of symmetric cryptographic primitives to evaluate the security is...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Doctor of Philosophy |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/160785 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Information security has received more and more attentions in recent decades with the rapid developments of the internet era. Since symmetric cryptographic primitives are widely used in current information systems, doing cryptanalysis of symmetric cryptographic primitives to evaluate the security is becoming increasingly significant. This thesis focuses on the cryptanalysis of block ciphers and hash functions assisted by tools including automatic tools and machine learning techniques, and shows the advantages of machine learning-aided and SAT-aided cryptanalysis over pure classical cryptanalysis. Firstly, regarding Keccak-f is the permutation used in the NIST SHA-3 hash function standard, we introduce a classical algorithm to exhaustively search for 3-round trail cores of Keccak-f [1600]. Then we develop a SAT-based automatic search toolkit to obtain differential trails for Keccak-f. With the help of this tool, we present the first 6-round classical collision attack on SHAKE128. Besides, we explore using neural networks to assist classical cryptanalysis and present the first practical 13-round neural-distinguisher-based key-recovery attacks on Speck32/64, which is a lightweight block cipher designed by NSA. |
|---|