20.500.12556/DKUM-88805 Implementacija in izvedba napadov ugibanja gesel na visoko zmogljivem računalniku diplomsko delo Implementation and execution of password guessing attacks on a high-performance computer Vdori v sisteme dandanes postajajo vedno pogostejši pojav. Ukradena gesla so pogosto prodana na internetu. Pomembno je, da gesla s primernimi metodami zaščitimo. V diplomski nalogi predstavljamo različne metode shranjevanja gesel in napadov nanja z uporabo visoko zmogljivih računalnikov. Poglobimo se v napade na obliko shranjenih gesel, kot so napad z grobo silo, napad z masko, slovarski napad, Markove verige in napad z mavrično tabelo. Predstavimo in obrazložimo avtorski program, ki implementira napade in njegove optimizacije. Program poženemo na HPC-ju in predstavimo rezultate zagona. Z analizo performančnih testov ugotovimo, da lahko dosežemo hitrejše izvajanje programa z boljšim poznavanjem ogrodja MPI, saj so določeni klici v našem primeru časovno potratnejši kot pričakovano. Ozdivni čas MPI funkcij se po določenemu številu iteracij poveča za več kot $100000\%$. Z vsakim dodatnim grozdom se čas izvajanja še dodatno poveča. Implementacija je v hitrosti na enem CPU-ju primerljiva s Hashcat-om. Boljše obvladovanje knjižnice MPI doprinese k bistveno hitrejšemu izvajanju programa. System intrusions are becoming an increasingly common occurance. Stolen passwords can be found all over the web, so it's more important than ever to protect password database with sufficiently strong methods. In this thesis we list different effective and ineffective methods of password storage. Viable attacks against hashed passwords, such as brute-force, dictionary, mask attacks, Markov chain attacks and rainbow table attacks are presented. Some of the aforementioned attacks are implemented using Rust against the popular hashing function SHA-256. The program is compiled and ran on an HPC. It's performance and results are graphed and explained. Analysis of those results shows, that performance can be improved with a better understanding of MPI. Execution of MPI library calls was slower than expected. The execution time of these MPI calls increases by over $100000\%$. With each extra node the execution time further increases. We can reach an acceptable multi-node performance with this program by improving MPI performance, as it rivals Hashcat when run on a single node. HPC gesla brute-force Rust slovar HPC password brute-force Rust dictionary true true false [A. Škof] Slovenski jezik Angleški jezik Diplomsko delo/naloga 2024-05-27 02:54:57 2024-07-01 11:44:31 2024-08-28 03:18:35 0000-00-00 00:00:00 2024 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (IX, 97 f.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.056.52(043.2) 205359619 UN_Skof_Aljaz_2024.pdf UN_Skof_Aljaz_2024.pdf 1 1323677E6AEB5CCE671D585475371D16 ceedb0029dcb67761f59230d1f8b7ad7a583b036ff6d8895e04c4c3ad70cd0c6 5b6421fe-1db6-11ef-8334-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=179885 UN_Skof_Aljaz_2024.zip UN_Skof_Aljaz_2024.zip 1 B7C87424820275502A59C7867D5360A0 7aed7c9a1d1dbbe08cce092d47d3b20c649e3613e3254a0a1cb24cc7cc311b2f 82f92f5d-1cf4-11ef-bd8c-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=179849 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0