In cryptography, acoustic cryptanalysis is a type of side channel attack which exploits sounds emitted by computers or machines. Modern acoustic cryptanalysis mostly focuses on the sounds produced by computer keyboards and internal computer components, but historically it has also been applied to impact printers and electromechanical cipher machines.
Victor Marchetti and John D. Marks eventually negotiated the declassification of CIA acoustic intercepts of the sounds of cleartext printing from encryption machines. Technically this method of attack dates to the time of FFT hardware being cheap enough to perform the task—in this case the late 1960s to mid-1970s. However, using other more primitive means such acoustical attacks were made in the mid-1950s.
In 2004, Dmitri Asonov and Rakesh Agrawal of the IBM Almaden Research Center announced that computer keyboards and keypads used on telephones and automated teller machines (ATMs) are vulnerable to attacks based on the sounds produced by different keys. Their attack employed a neural network to recognize the key being pressed. By analyzing recorded sounds, they were able to recover the text of data being entered. These techniques allow an attacker using covert listening devices to obtain passwords, passphrases, personal identification numbers (PINs), and other information entered via keyboards.
In 2005, a group of UC Berkeley researchers performed a number of practical experiments demonstrating the validity of this kind of threat.
Also in 2004, Adi Shamir and Eran Tromer demonstrated that it may be possible to conduct timing attacks against a CPU performing cryptographic operations by analysis of variations in humming emissions (that is, its ultrasonic noise emanating from capacitors on a motherboard, not electromagnetic emissions or the human-audible humming of a cooling fan).
This kind of cryptanalysis can be defeated by generating sounds that are in the same spectrum and same form as keypresses. If you randomly replay sounds of actual keypresses, it may be possible to totally defeat such kinds of attacks. It is advisable to use at least 5 different recorded variations (36 x 5 = 180 variations) for each keypress to get around the issue of FFT fingerprinting. Alternatively, white noise of a sufficient volume (which may be simpler to generate for playback) will also mask the acoustic emanations of individual keypresses.
- Marchetti, Victor; Marks, John (1973), The CIA and the Craft of Intelligence
- Wright, Peter (1987), Spycatcher: The candid autobiography of a senior intelligence officer, Viking
- Yang, Sarah (14 September 2005), "Researchers recover typed text using audio recording of keystrokes", UC Berkeley News
- Shamir, Adi; Tromer, Eran, Acoustic cryptanalysis: On nosy people and noisy machines
- Adi Shamir & Eran Tromer. "Acoustic cryptanalysis". Blavatnik School of Computer Science, Tel Aviv University. Retrieved 1 November 2011.
- Asonov, Dmitri; Agrawal, Rakesh (2004), "Keyboard Acoustic Emanations" (PDF), IBM Almaden Research Center