Can a Feistel cipher be accurately evaluated by a third party?

  • Thread starter Thread starter Chris Miller
  • Start date Start date
Click For Summary
SUMMARY

The forum discussion centers on the evaluation of a newly developed block cipher based on a Feistel network, which is claimed to be smaller and faster than AES (Rijndael) by 15 to 40 percent. The cipher utilizes a customizable s-box, allowing for various configurations and random sequences. The author seeks advice on how to have the cipher examined by experienced cryptographers, suggesting a challenge for decryption as a means of validation. However, another participant expresses skepticism about the feasibility of decrypting the cipher without advanced attack capabilities, such as chosen plaintext or chosen ciphertext attacks.

PREREQUISITES
  • Understanding of Feistel networks and their structure
  • Familiarity with AES (Rijndael) and its operational principles
  • Knowledge of cryptographic attack methods, including chosen plaintext and chosen ciphertext attacks
  • Experience with cryptographic s-boxes and their configurations
NEXT STEPS
  • Research the design principles of Feistel ciphers and their security implications
  • Learn about AES (Rijndael) performance metrics and comparisons with other ciphers
  • Study chosen plaintext and chosen ciphertext attack strategies in cryptography
  • Explore methods for evaluating and validating cryptographic algorithms
USEFUL FOR

Cryptographers, security researchers, and developers interested in evaluating or creating block ciphers, particularly those exploring Feistel networks and their performance compared to established standards like AES.

Chris Miller
Messages
371
Reaction score
35
I've developed a block cipher based on a Feistel network that is, round for round, smaller (4 steps) and faster (by 15 to 40 percent) than AES (aka Rijndael), and way simpler. I'd like to say stronger, too, but cannot support this assertion myself. Proprietary configurations are easy to create. Because it's a Feistel network (vs. a Substitution Permutation Network like AES), the s-box (an expansion p-box actually) , which is a table of 256 (or 1024 or 2048 in some flavors) 32-bit values, can be populated with virtually any normal random sequence, or even scheduled as part of the key. As with AES, it exhibits BIC and SAC compliance in two rounds.

My question is, is there any way of having it examined, if only cursorily, by someone better educated and experienced?
 
Computer science news on Phys.org
Post the source code on a website, social network, or programming site and a challenge of decrypting a message that you supply (maybe a random paragraph from your favorite novel). 'Reward' can be anything from a 'Thank You', to acknowledgment in the distribution version, a custom T-shirt, or perhaps $50 to the first correct decode.

Cheers,
Tom
 
Thanks Tom, but I am 100 percent sure no one is going to decrypt the cipertext of even a mediocre 128-bit multi-round Feistel cipher without at least the ability to mount a chosen plaintext or chosen ciphertext attack (i.e., oracle access). For me to invite someone to try strikes me as a little naive. Like just providing 1 block (16 bytes) of ciphertext, a simple XOR of the plaintext message with the key (i.e., a one-time pad) would be enough to prevent decryption.