Technical Article

EM Side-Channel Attacks on Cryptography

July 26, 2023 by Jake Hertz

Electromagnetic-based side-channel attacks are non-invasive, meaning the attacker does not need physical access to the device to steal information. We’ll look at how these EM side-channel attacks work.

We previously introduced the concept of side-channel attacks: what they are and why they are significant hardware security threats. Of the many forms of side-channel attacks, one of the most powerful is electromagnetic (EM) attacks.

In this article, we’ll describe EM attacks, how they work, and take a look at specific forms of power attacks that have been used to break cryptography in the past.


What Are EM Attacks and Why Are They Powerful?

An EM side-channel attack is a form of attack that exploits the electromagnetic emanations from an electronic device as a form of information leakage (Figure 1). EM attacks have been studied and researched for decades and, as such, have become some of the most well-understood and powerful hardware attacks. These attacks offer unique features that make them very powerful relative to other side-channel attacks.


Example setup for an EM-based attack

Figure 1. Example setup for an EM-based attack. Image used courtesy of Das, et al.


EM Attacks are Non-Invasive

Most importantly, EM attacks are non-invasive, meaning that the attacker does not need to physically access the device in order to perform the attack. First, this makes EM attacks powerful because they are easy for the attacker to perform. It only requires the use of a near-field probe and an oscilloscope.

Another corollary of this fact is that EM attacks don’t require any modification to the device under attack. Many power-based attacks require specialized tools or IC decapsulation.


EM Attacks are Difficult to Detect

Beyond this, the non-invasive nature of EM attacks means that it is virtually impossible for the victims to recognize that they have been attacked. This further makes EM attacks powerful, as they are difficult to identify and stop.


EM Attacks Improve Data Collection

Finally, EM-based attacks have been shown to yield data with a higher signal-to-noise ratio (SNR) than power attacks. Therefore, the attacks require less signal collection to filter out the noise.


How EM Attacks Work

As defined by Faraday’s Law, electric currents generate a corresponding magnetic field. An EM-based side-channel attack leverages this fact by monitoring the EM radiation emitted from a device during operation to steal information. As illustrated conceptually in Figure 2, EM-based attacks physically measure the electromagnetic emanations from an electronic device and use analytical methods and leakage models to steal information from the data.


EM-based attacks are used to extract information from the data

Figure 2. EM-based attacks are used to extract information from the data. Image used courtesy of


In CMOS devices, the current flow occurs primarily when there is a change in the logic state at a clock edge. In digital logic, these changing states, and hence their current and EM emanations, are related to binary bit streams of data.

This means there can be a direct deterministic relationship between EM emanations in a CMOS device and data being processed within the device. This is the point of an EM attack, to steal information from the relationship between EM radiations and the device’s current draw.

As shown in Figure 3, an attacker can use an EM probe, generally tuned for a device’s fundamental frequency and its harmonics, to capture these traces. Then, various statistical methods and leakage models are employed to extract sensitive information from the data.


Example setup for an EM-based attack

Figure 3. Example setup for an EM-based attack. Image used courtesy of Das, et al.


Types of EM Side-Channel Attacks

Similar to power attacks, EM attacks can be divided into two main categories:

  1. Simple Electromagnetic Analysis (SEMA)
  2. Differential Electromagnetic Analysis (DEMA)

The primary difference between the two is that, in SEMA, attackers try to interpret the data traces directly, while in DEMA, attackers collect large numbers of traces and run differential statistical methods on the data to identify data-dependent correlations. DEMA attacks are undoubtedly more robust, powerful, and widely applicable, but they also tend to be more complicated and time-consuming.

Today, analysis methods utilizing Machine Learning pattern recognition and classification are becoming increasingly popular as Machine Learning becomes increasingly accessible and sophisticated.

EM attacks have been proven to successfully steal sensitive information from a variety of cryptographic devices. As a testament to the strength of EM attacks, Fox IT used an EM attack to break an AES-256 crypto core in just five minutes while at a distance of 1 m from the device.

Beyond this, other research has shown the power of EM attacks on IoT devices. In a paper by Syakkara, et al., entitled Leveraging Electromagnetic Side-Channel Analysis for the Investigation of IoT Devices, researchers were able to use EM attacks to non-invasively detect what cryptographic algorithm an IoT device is running, what software program it is running, and what version of firmware the device is hosting using the setup demonstrated above in Figure 1.


Electronic Devices Vulnerable to EM Attacks

Another aspect of what makes EM attacks so powerful is the broad range of electronic devices that they have the potential to affect. Some of the devices that have been shown to be susceptible to EM-based side-channel attacks include:

  • Smart Cards
  • Mobile Payment Systems
  • Embedded Devices
  • Smartphones
  • IoT Devices
  • SoCs


While this list is not exhaustive, it is telling—almost all of the electronic devices we use in our daily lives are susceptible to EM attacks in one way or another.


EM Attack Takeaways

Hopefully, this article and deep dive into a specific set of attacks done in academia provide insight into just how powerful EM attacks can be. EM attacks, especially when paired with tools like machine learning, can pose significant threats to electronic devices if the devices are not properly prepared for such threats.