Today I will show you how I can exfiltrate some secret data from a machine which is well isolated. This particular machine has no network connectivity, the whole interface is being disconnected. The idea of how to exfiltrate the data is relatively simple: if I have my machine there is some data

over here, I will convert the data into a Doom wave file for the sound, and then I will play this sound through the speaker. It will be recorded on the other end by some mic, and the wave file will be created from the recorded sound and it will be converted

back to the secret data effectively. The secret data is being exfiltrated to some other machine. So to create a wave file, first I have a Python script. If your machine is totally disconnected, you have to write it down from scratch. I'm defining how long is one bit

what our frequency is, from 0 and 1 bit. I'm reading a file then creating a wave file, filling it up with bits bit by bit for every single bit of my secret data. And that's basically it for creating the file itself. So I will run

the script and my resulting data is created over here. It is ready to be played. For some tricky reason, I will show you later on why I will play it twice at the same time. Obviously, I have to record it. I will use Audacity software, an open source software. It will be noisy

so I will start recording and I will play my file. Foreign, as I said, I have played it twice because if you observe the way how it is played, the first one does not start very sharply, which I like, but the second one is, is a small trick for me. It is good enough to

have my data in slightly better quality, so it is a recorded file. Everything before this point is useless for me, so shift Home, Dell, and now I can also delete something that is over here. So, everything after this point, let's delete it, so I have a track

containing the frequencies being recorded, containing the information about my Beats. Now, I will export it into Dome, a wave file. I'm using an uncompressed wave file at 44,100 kilobytes and a single audio track, mono audio. I will save it. Yes, I want to replace it, and now

I have my output recorded wave being created right now. I have to decode it back to the data. So, I'm providing the same information. I'm using here the Gertzel function. The Gertzel function is a function, is an algorithm, answering me for the question, if a frequency K

appears within my data set, dollar a year, in my case. So, what I'm doing actually, I'm reading the wave file into the set of bits, into an array, and then I'm cutting the header, and then I'm going through the um file, analyzing what, how 1,000 of samples

from each bit determined by the time within the wave file, and if I have one bit or zero bit, I'm assembling them into a byte. If I have a complete byte, I'm displaying it, and that's basically, I can also save it to a file. If I run it on my

WAV file being recorded, F5, I can see the set of bytes being decoded, and obviously, and it is exactly the same set of bytes I had within my uh, the data file I have exfiltrated. Even though if you want to play on your own, feel free. I'm

publishing it on my PSPs repository under no DLP for older, so feel free to modify those PowerShell scripts and play around on your own, etc., etc. That's basically it.

Exfiltrating the data through audio

Grzegorz Tworek

유출 방법: 데이터 변환 및 오디오 전송

네트워크 연결이 없는 격리된 머신에서 비밀 데이터 유출.

오디오 전송을 위해 데이터를 둠 웨이브 파일로 변환합니다.

비트 길이와 주파수를 정의하는 파이썬 스크립트로 웨이브 파일을 생성합니다.

스크립트 실행 및 웨이브 파일 생성.

Audacity 소프트웨어를 사용한 오디오 녹음, 더 나은 음질을 위해 파일을 두 번 재생합니다.

오디오 파일 처리 및 데이터 추출

오디오 파일 정리 및 WAV 파일로 내보내기(비압축, 44,100kbps, 모노).

거첼 함수를 사용하여 WAV 파일을 디코딩하여 주파수를 식별합니다.

웨이브 파일을 읽고, 샘플을 분석하고, 비트를 바이트로 조합합니다.

원본 데이터와 일치하는 성공적인 데이터 복구.

'이전 버전에 대한 DLP 없음' 아래의 PSP 리포지토리에서 코드 가용성.