• narrowband harmonics of the digital clock signals ;
• broadband harmonics of the various 'random' digital signals such as the video signal.^[2]

Known as compromising emanations or TEMPEST radiation, a code

Known as compromising emanations or TEMPEST radiation, a code word for a U.S. government programme aimed at attacking the problem, the electromagnetic broadcast of data has been a significant concern in sensitive computer applications. Eavesdroppers can reconstruct video screen content from radio frequency emanations.^[3] Each (radiated) harmonic of the video signal shows a remarkable resemblance to a broadcast TV signal. It is therefore possible to reconstruct the picture displayed on the video display unit from the radiated emission by means of a normal television receiver.^[2] If no preventive measures are taken, eavesdropping on a video display unit is possible at distances up to several hundreds of meters, using only a normal black-and-white TV receiver, a directional antenna and an antenna amplifier. It is even possible to pick up information from some types of video display units at a distance of over 1 kilometer. If more sophisticated receiving and decoding equipment is used, the maximum distance can be much greater.^[4]

Compromising reflections

When a key is pressed, released or held down, the keyboar

The four different kinds way of compromising electromagnetic emanations are described below.

When a key is pressed, released or held down, the keyboard sends a packet of information known as a scan code to the computer.^[13] The protocol used to transmit these scan codes is a bidirectional serial communication, based on four wires: Vcc (5 volts), ground, data and clock.^[13] Clock and data signals are identically generated. Hence, the compromising emanation detected is the combination of both signals. However, the edges of the data and the clock lines are not superposed. Thus, they can be easily separated to obtain independent signals.^[14]

The Generalized Transition Technique

The Falling Edge Transition attack is limited to a partial recovery of the keystrokes. This is a significant limitation.^[15] The GTT is a falling edge transition attack improved, which recover almost all keystrokes. Indeed, between two traces, there is exactly one data rising edge. If attackers are able to detect this transition, they can fully recover the keystrokes.^[15]

The Modulation Technique

Harmonics compromising electromag

Harmonics compromising electromagnetic emissions come from unintentional emanations such as radiations emitted by the clock, non-linear elements, crosstalk, ground pollution, etc. Determining theoretically the reasons of these compromising radiations is a very complex task.^[16] These harmonics correspond to a carrier of approximately 4 MHz which is very likely the internal clock of the micro-controller inside the keyboard. These harmonics are correlated with both clock and data signals, which describe modulated signals (in amplitude and frequency) and the full state of both clock and data signals. This means that the scan code can be completely recovered from these harmonics.^[16]

The Matrix Scan TechniqueKeyboard manufacturers arrange the keys in a matrix. The keyboard controller, often an 8-bit processor, parses columns one-by-one and recovers the state of 8 keys at once. This matrix scan process can be described as 192 keys (some keys may not be used, for instance modern keyboards use 104/105 keys) arranged in 24 columns and 8 rows.^[17] These columns are continuously pulsed one-by-one for at least 3μs. Thus, these leads may act as an antenna and generate electromagnetic emanations. If an attacker is able to capture these emanations, he can easily recover the column of the pressed key. Even if this signal does not fully describe the pressed key, it still gives partial information on the transmitted scan code, i.e. the column number.^[17]

Note that the matrix scan routine loops continuously. When no key is pressed, we still have a signal composed of multiple equidistant peaks. These emanations may be used to remotely detect the presence of powered computers. Concerning wireless keyboards, the wireless data burst transmission can be used as an electromagnetic trigger to detect exactly when a key is pressed, while the matr

Note that the matrix scan routine loops continuously. When no key is pressed, we still have a signal composed of multiple equidistant peaks. These emanations may be used to remotely detect the presence of powered computers. Concerning wireless keyboards, the wireless data burst transmission can be used as an electromagnetic trigger to detect exactly when a key is pressed, while the matrix scan emanations are used to determine the column it belongs to.^[17]

Some techniques can only target some keyboards. This table sums up which technique could be used to find keystroke for different kind of keyboard.

Technique name	Wired Keyboard	Laptop Keyboard	Wireless Keyboard
Falling Edge Transition Technique	Yes	Yes
Generalized Transition Technique	Yes	Yes
Modu In their paper called "Compromising Electromagnetic Emanations of Wired and Wireless Keyboards", Martin Vuagnoux and Sylvain Pasini tested 12 different keyboard models, with PS/2, USB connectors and wireless communication in different setups: a semi-anechoic chamber, a small office, an adjacent office and a flat in a building. The table below presents their results. Attacks against emanations caused by human typing have attracted interest in recent years. In particular, works showed that keyboard acoustic emanations do leak information that can be exploited to reconstruct the typed text.[18] PC keyboards, notebook keyboards are vulnerable to attacks based on differentiating the sound emanated by different keys.[19] This attack takes as input an audio signal containing a recording of a single word typed by a single person on a keyboard, and a dictionary of words. It is assumed that the typed word is present in the dictionary. The aim of the attack is to reconstruct the original word from the signal.[20] This attack, taking as input a 10-minute sound recording of a user typing English text using a keyboard, and then recovering up to 96% of typed characters.[21] This PC keyboards, notebook keyboards are vulnerable to attacks based on differentiating the sound emanated by different keys.[19] This attack takes as input an audio signal containing a recording of a single word typed by a single person on a keyboard, and a dictionary of words. It is assumed that the typed word is present in the dictionary. The aim of the attack is to reconstruct the original word from the signal.[20] This attack, taking as input a 10-minute sound recording of a user typing English text using a keyboard, and then recovering up to 96% of typed characters.[21] This attack is inexpensive because the other hardware required is a parabolic microphone and non-invasive because it does not require physical intrusion into the system. The attack employs a neural network to recognize the key being pressed.[19] It combines signal processing and efficient data structures and algorithms, to successfully reconstruct single words of 7-13 characters from a recording of the clicks made when typing them on a keyboard.[18] The sound of clicks can differ slightly from key to key, because the keys are positioned at different positions on the keyboard plate, although the clicks of different keys sound similar to the human ear.[19] On average, there were only 0.5 incorrect recognitions per 20 clicks, which shows the exposure of keyboard to the eavesdropping using this attack.[22] The attack is very efficient, taking under 20 seconds per word on a standard PC. A 90% or better success rate of finding the correct word for words of 10 or more characters, and a success rate of 73% over all the words tested.[18] In practice, a human attacker can typically determine if text is random. An attacker can also identify occasions when the user types user names and passwords.[23] Short audio signals containing a single word, with seven or more characters long was considered. This means that the signal is only a few seconds long. Such short words are often chosen as a password.[18] The dominant factors affecting the attack's success are the word length, and more importantly, the number of repeated characters within the word.[18] This is a procedure that makes it possible to efficiently uncover a word out of audio recordings of keyboard click sounds.[24] More recently, extracting information out of another type of emanations was demonstrated: acoustic emanations from mechanical devices such as dot-matrix printers.[18] While extracting private information by watching somebody typing on a keyboard might seem to be an easy task, it becomes extremely challenging if it has to be automated. However, an automated tool is needed in the case of long-lasting surveillance procedures or long user activity, as a human being is able to reconstruct only a few characters per minute. The paper "ClearShot: Eavesdropping on Keyboard Input from Video" presents a novel approach to automatically recovering the text being typed on a keyboard, based solely on a video of the user typing.[25] Automatically recognizing the keys being pressed by a user is a hard problem that requires sophisticated motion analysis. Experiments show that, for a human, reconstructing a few sentences requires lengthy hours of slow-motion analysis of the video.[26] The attacker might install a surveillance device in the r Automatically recognizing the keys being pressed by a user is a hard problem that requires sophisticated motion analysis. Experiments show that, for a human, reconstructing a few sentences requires lengthy hours of slow-motion analysis of the video.[26] The attacker might install a surveillance device in the room of the victim, might take control of an existing camera by exploiting a vulnerability in the camera's control software, or might simply point a mobile phone with an integrated camera at the laptop's keyboard when the victim is working in a public space.[26] Balzarotti's analysis is divided into two main phases (figure below). The first phase analyzes the video recorded by the camera using computer vision techniques. For each frame of the video, the computer vision analysis computes the set of keys that were likely pressed, the set of keys that were certainly not pressed, and the position of space characters. Because the results of this phase of the analysis are noisy, a second phase, called the text analysis, is required. The goal of this phase is to remove errors using both language and context-sensitive techniques. The result of this phase is the reconstructed text, where each word is represented by a list of possible candidates, ranked by likelihood.[26] With acoustic emanations, an attack that recovers what a dot-matrix printer processing English text is printing is possible. It is based on a record of the sound the printer makes, if the microphone is close enough to it. This attack recovers up to 72% of printed words, and up to 95% if knowledge about the text are done, with a microphone at a distance of 10 cm from the printer.[27] After an upfront training phase ("a" in the picture below), the attack ("b" in the picture below) is fully automated and uses a combination of machine learning, audio processing, and speech recognition techniques, including spectrum features, Hidden Markov Models and linear classification.[5] The fundamental reason why the reconstruction of the printed text works is that, the emitted sound becomes louder if more needles strike the paper at a given time.[9] There is a correlation between the number of needles and the intensity of the acoustic emanation.[9] A training phase was conducted where words from a dictionary are printed and characteristic sound features of these words are extracted and stored in a database. The trained characteristic features was used to recognize the printed English text.[9] But, this task is not trivial. Major challenges inclu After an upfront training phase ("a" in the picture below), the attack ("b" in the picture below) is fully automated and uses a combination of machine learning, audio processing, and speech recognition techniques, including spectrum features, Hidden Markov Models and linear classification.[5] The fundamental reason why the reconstruction of the printed text works is that, the emitted sound becomes louder if more needles strike the paper at a given time.[9] There is a correlation between the number of needles and the intensity of the acoustic emanation.[9] A training phase was conducted where words from a dictionary are printed and characteristic sound features of these words are extracted and stored in a database. The trained characteristic features was used to recognize the printed English text.[9] But, this task is not trivial. Major challenges include : Timing attacks enable an attacker to extract secrets maintained in a security system by observing the time it takes the system to respond to various queries.[28] SSH is designed to provide a secure channel between two hosts. Despite the encryption and authentication mechanisms it uses, SSH has weaknesses. In interactive mode, every individual keystroke that a user types is sent to the remote machine in a separate IP packet immediately after the key is pressed, which leaks the inter-keystroke timing information of users’ typing. Below, the picture represents the command su processed through a SSH connection. Network messages sent between the host and the client for the command 'su' – numbers are size of network packet in byte A very simple statistical techniques suffice to reveal sensitive information such as the length of users’ passwords or even root passwords. By using advanced statistical techniques on timing information collected from the network, the eavesdropper can learn significant information about what users type in SSH sessions.[29] Because the time it takes the operating system to send out the packet after the keypress is in general negligible comparing to the interkeystroke timing, this also enables an eavesdropper to learn the precise interkeystroke timings of users’ typing from the arrival times of packets.[30] SSH is designed to provide a secure channel between two hosts. Despite the encryption and authentication mechanisms it uses, SSH has weaknesses. In interactive mode, every individual keystroke that a user types is sent to the remote machine in a separate IP packet immediately after the key is pressed, which leaks the inter-keystroke timing information of users’ typing. Below, the picture represents the command su processed through a SSH connection. A very simple statistical techniques suffice to reveal sensitive information such as the length of users’ passwords or even root passwords. By using advanced statistical techniques on timing information collected from the network, the eavesdropper can learn significant information about what users type in SSH sessions.[29] Because the time it takes the operating system to send out the packet after the keypress is in general negligible comparing to the interkeystroke timing, this also enables an eavesdropper to learn the precise interkeystroke timings of users’ typing from the arrival times of packets.[30] Memory Physical chemistry Data remanence problems not only affect obvious areas such as RAM and non-volatile memory cells but can also occur in other areas of t Data remanence problems not only affect obvious areas such as RAM and non-volatile memory cells but can also occur in other areas of the device through hot-carrier effects (which change the characteristics of the semiconductors in the device) and various other effects which are examined alongside the more obvious memory-cell remanence problems.[31] It is possible to analyse and recover data from these cells and from semiconductor devices in general long after it should (in theory) have vanished.[32] Electromigration, which means to physically move the atom to new locations (to physically alter the device itself) is another type of attack.[31] It involves the relocation of metal atoms due to high current densities, a phenomenon in which atoms are carried along by an "electron wind" in the opposite direction to the conventional current, producing voids at the negative electrode and hillocks and whiskers at the positive electrode. Void formation leads to a local increase in current density and Joule heating (the interaction of electrons and metal ions to produce thermal energy), producing further el Electromigration, which means to physically move the atom to new locations (to physically alter the device itself) is another type of attack.[31] It involves the relocation of metal atoms due to high current densities, a phenomenon in which atoms are carried along by an "electron wind" in the opposite direction to the conventional current, producing voids at the negative electrode and hillocks and whiskers at the positive electrode. Void formation leads to a local increase in current density and Joule heating (the interaction of electrons and metal ions to produce thermal energy), producing further electromigration effects. When the external stress is removed, the disturbed system tends to relax back to its original equilibrium state, resulting in a backflow which heals some of the electromigration damage. In the long term though, this can cause device failure, but in less extreme cases it simply serves to alter a device's operating characteristics in noticeable ways. For example, the excavations of voids leads to increased wiring resistance and the growth of whiskers leads to contact formation and current leakage.[33] An example of a conductor which exhibits whisker growth due to electromigration is shown in the figure below: One example which exhibits void formation (in this case severe enough to have led to complete failure) is shown in this figure: Void formation due to electromigration Contrary to popular assumption, DRAMs used in most modern computers retain their contents for several seconds after power is lost, even at room temperature and even if removed from a motherboard.[34] Many products do cryptographic and other security-related computations using secret keys or other variables that the equipment's operator must not be able to read out or alter. The usual solution is for the secret data to be kept in volatile memory inside a tamper-sensing enclosure. Security processors typically store secret key material in static RAM, from which power is removed if the device is tampered with. At temperatures below −20 °C, the contents of SRAM can be ‘frozen’. It is interesting to know the period of time for which a static RAM device will retain data once the power has been removed. Low temperatures can increase the data retention time of SRAM to many seconds or even minutes.[35] Many products do cryptographic and other security-related computations using secret keys or other variables that the equipment's operator must not be able to read out or alter. The usual solution is for the secret data to be kept in volatile memory inside a tamper-sensing enclosure. Security processors typically store secret key material in static RAM, from which power is removed if the device is tampered with. At temperatures below −20 °C, the contents of SRAM can be ‘frozen’. It is interesting to know the period of time for which a static RAM device will retain data once the power has been removed. Low temperatures can increase the data retention time of SRAM to many seconds or even minutes.[35] Maximillian Dornseif presented a technique in these slides, which let him take the control of an Apple computer thanks to an iPod. The attacks needed a first generic phase where the iPod software was modified so that it behaves as master on the FireWire bus. Then the iPod had full read/write access on the Apple Computer when the iPod was plugged into a FireWire port.[36] FireWire is used by : audio devices, printers, scanners, cameras, gps, etc. Generally, a device connected by FireWire has full access (read/write). Indeed, OHCI Standard (FireWire standard) reads : Physical requests, including physical read, physical write and lock requests to some CSR registers (section 5.5), are handled directly by the Host Controller without assistance by system softwar Physical requests, including physical read, physical write and lock requests to some CSR registers (section 5.5), are handled directly by the Host Controller without assistance by system software. — OHCI Standard So, any device connected by FireWire can read and write data on the computer memory. For example, a device can : Grab the screen contents ; Just search the memory for stri So, any device connected by FireWire can read and write data on the computer memory. For example, a device can : Grab the screen contents ; Just search the memory for strings su or Mess up the memory ; Change screen content ; Change UID/GID of a certain process ; Inject code into a process ; Inject an additional process. Processor< To increase the computational power, processors are generally equipped with a cache memory which decreases the memory access latency. Below, the figure shows the hierarchy between the processor and the memory. First the processor looks for data in the cache L1, then L2, then in the memory. Processor cache hierarchy When the data is not where the processor is looking for, it is called a cache-miss. Below, pictures show how the processor fetch data when there are two cache levels. Type of keyboard Number of tested keyboard FETT GTT MT MST PS/2 7 7/7 6/7 4/7 5/7 USB 2 0/2 0/2 0/2 2/2 Laptop 2 1/2 1/2 0/2 2/2 Wireless 1 0/1 0/1 0/1 [37] The basic idea proposed in this paper is to force a cache miss while the processor is executing the AES encryption algorithm on a known plain text.[38] The attacks allow an unprivileged process to attack other process running in parallel on the same processor, despite partitioning methods such as memory protection, sandboxing and virtualization.[39] Timing attack By carefully measuring the amount of time required to perform private key operations, attackers may be able to find fixed Diffie-Hellman exponents, factor RSA keys, and break other cryptosystems. Against a vulnerable system, the attack is computationally inexpensive and often requires only known ciphertext.[40] The attack can be treated as a signal detection problem. The signal consists of the timing variation due to the target exponent bit, and noise results from measurement inaccuracies and timing variations due to unknown exponent bits. The properties of the signal and noise determine the number of timing measurements required to for the attack. Timing attacks can potentially be used against other cryptosystems, including symmetric functions.[41] Privilege escalation Diffie-Hellman exponents, factor RSA keys, and break other cryptosystems. Against a vulnerable system, the attack is computationally inexpensive and often requires only known ciphertext.[40] The attack can be treated as a signal detection problem. The signal consists of the timing variation due to the target exponent bit, and noise results from measurement inaccuracies and timing variations due to unknown exponent bits. The properties of the signal and noise determine the number of timing measurements required to for the attack. Timing attacks can potentially be used against other cryptosystems, including symmetric functions.[41] Privilege escalation