This was before the era of iPods and SDCard players, so I had my mobile phone in a setup where I would call back another phone connected to my Pentium MMX 233MHz at home, that ran a sort of audio directory that would playback a certain lecture recording I would select from the menu, using DTMF tones.
I had a small keyboard sewn into my sleeve that connected to the customized mobile phone via a DB9 plug and then to the numeric keyboard, allowing DTMF codes to be sent by gently and invisibly moving my fingers. The whole setup was hideous and had it's own dedicated jacket, with wires, phone, keyboard, audio amplifier, neckloop, the earphone... a complete cyborg for academic fraud.
Back on the PC side, I wrote a C++ application from scratch that would capture audio via the soundcard using Windows Wave API, decode the DTMF pulses using a couple of IIR filters then navigate the menu and playback the required file to the mobile phone connected to the soundcard. The C++ program and menu system was scripted using an .INI file that defined the structure with links to various ADPCM-compressed .wav files that represented the menu headings or the leaf content itself (a good structure was necessary to quickly access the correct lecture after receiving the exam subjects).
Work-wise, it was a lot more difficult then putting the effort in memorizing the stuff, but I rejected memorisation on principle, that's not what an university should be about. The whole thing turned out to be a massive learning project, but I obviously could not speak about it at interviews. It's the first time I mention it to anybody.
I used the setup for two exams that I aced, was never caught but it was nerve-wracking to use in close proximity to a teacher.