Fast Transciphering Via Batched And Reconfigurable LUT Evaluation

This repository contains the materials necessary to replicate the timings of the article:

“Fast Transciphering Via Batched And Reconfigurable LUT Evaluation”. IACR Transactions on Cryptographic Hardware and Embedded Systems, vol. 2024, no. 4, Sept. 2024, pp. 205-30, https://doi.org/10.46586/tches.v2024.i4.205-230.

Build

Build instructions:

• Install Clang Version >= 15 and set the CC and CXX variables accordingly.
• chmod +x install_openfhe.sh
• ./install_openfhe.sh 64. Replace 64 by 32 for Paramter Set I or II, and change the set as well in LUTEvalParams.h when using a set different from Set IV
• mkdir build && cd build
• cmake ..
• make -j 16

Execution

For optimal performance during execution, we suggest the following:

• Important, especially on laptops. Determine which CPU cores are performance cores (as opposed to E cores), e.g. via a system monitor
• Use cpufreq-set to the the cpu frequency to maximum. By default, processors will not run at the maximum frequency.
• Several example_* should be in the build directory. The most important binaries are example_aes and example_bitdecomp (for LUT timing)
• Execute using taskset and assign the core which was set to maximum frequency

Related works

The folder related_work_scripts contains two scripts. The first one contains a sage script using the lattice estimator to estimate the security of the first related work on transciphering (TSBS23). The second script estimates the failure probability of related work number two (WWL24).