Paper 2025/1539

EvH: Randomized Symmetric Cipher Paradigm with Holographic Storage and Parallelism, Compression, & Erasure Recovery Integration

Abstract

Standard symmetric encryption schemes, such as AES, other block ciphers and their modes, and stream ciphers, are highly effective and efficient for many standard scenarios. All of them have been derived from Shannon's 1948 seminal work on the communication theory of secrecy systems. Here we look at other settings where the situation is somewhat different from the standard one: \textit{e}.\textit{g}., while the encryption process may fail to update the ciphertext a limited number of times, or bits of the ciphertext are omitted: can the decryption process nevertheless recover the message in its entirety? Another situation is when encrypting a bulk of messages that must be packed together within the same dedicated ciphertext space (i.e., encryption done holographically on a clockchain space). Can a process compress the messages this way? Another issue is adding a further hiding layer to ciphertexts, like hiding the number of messages packed together, or even attempting to hide that encryption (rather than another cryptographic protocol) takes place? Can the new paradigm be based directly on a simple cryptographic (preferably post-quantum) tool? Note that the above scenarios, involving data compression and correction, are naturally derived from Shannon's other 1948 seminal work on Information Theory.   This paper introduces Encryption via Hash (EvH), a A new symmetric randomized cipher built upon a pseudorandom keyed cryptographic hash (i.e., Message Authentication Code, MAC functions), and Bloom Filters. EvH's core novelty lies in its prefix decryption capability. This unique property enables a paradigm in which encryption is tightly integrated with online compression and robust resilience to omission errors of the encryption process. By representing message prefixes in a Bloom filter, EvH allows a receiver to decrypt the initial part of a message even if subsequent data are lost, and to recover from some prefix omissions during encryption. At times, these built-in new properties may be significant (even beyond the above examples). Furthermore, this prefix-based approach facilitates simultaneous compression during the decryption phase by dynamically pruning invalid message continuations, using shared $\mu$-gram dictionaries or employing search via Large Language Models (LLMs). The result is a stateless and parallelizable cipher that, while computationally distinct from traditional ciphers, offers unique functional benefits for specific use cases, at the cost of correctness being ensured only probabilistically (as in compression processes), though the error can be well controlled and made significantly small.