We show that PC keyboards, notebook keyboards, telephone and ATM pads are vulnerable to attacks based on differentiating the sound emanated by different keys. Our attack employs a neural network to recognize the key being pressed. We also investigate why different keys produce different sounds and provide hints for the design of homophonic keyboards that would be resistant to this type of attack. Experimental Set-up We first specify the equipment and the software used in our study. Keyboards.

more »

... used several types of keyboards. Most of PC keyboard experiments were performed with an IBM keyboard S/N 0953260, P/N 32P5100. Experiments with multiple keyboards were performed with three GE Power keyboards HO97798. For experiments with telephones, Siemens RP240 phones (M/N 62001) were used. Microphones. We used a simple PC microphone for short distances up to 1 meter and a parabolic microphone for eavesdropping from a distance. Computer omnidirection microphone: serial number 33 − 3026 manufactured by RadioShack; frequency response: 30 Hz-15 kHz; impedance: 1000 ohms ±30%; sensitivity: −68 dB ±3 dB; operating voltage: 1.0 to 10 VDC. Parabolic microphone: 'Bionic Booster' manufactured by Silver Creek Industries; frequency response 100 Hz-10 kHz (−3 dB response); gain amp. cut off at 90 dB; overall system gain: 40 dB; sensitivity: −46 dB (0 dB = 1 V/Pa). ADC and FFT. The input was digitized using a standard PC sound card with 44.1 kHz sampling rate. Sigview software v.1.81 was used for recording the sound as well as for calculating time-FFT on 2 ms windows, with the Hanning windowing function applied. Window overlap was not available.