Smartphone Encryption Cracked: A New Side-Channel Threat

Author: Denis Avetisyan

Researchers demonstrate practical electromagnetic attacks can bypass modern smartphone security, even on devices with secure enclaves.

The experimental setup details the measurement of $EMSCA$ specifically for the application target within the Fairphone 4 device.

This review details the challenges and feasibility of breaking ECDSA on contemporary smartphones using electromagnetic side-channel analysis, highlighting vulnerabilities beyond software-level defenses.

Despite increasing reliance on smartphones for sensitive transactions, the security of their cryptographic implementations against physical attacks remains surprisingly underexplored. This research, ‘Breaking ECDSA with Electromagnetic Side-Channel Attacks: Challenges and Practicality on Modern Smartphones’, assesses the feasibility of exploiting electromagnetic side-channel vulnerabilities in contemporary smartphone systems-on-chip. We demonstrate successful recovery of ECDSA secrets from OpenSSL on both Raspberry Pi 4 and Fairphone 4 devices, revealing that existing software countermeasures are insufficient to prevent practical attacks. Given the evolving landscape of mobile security-and the imminent deployment of applications like the European Digital Identity wallet-what robust hardware and software protections are needed to safeguard sensitive data on all smartphones?

The Expanding Attack Surface: Smartphones as Data Concentrators

The modern smartphone has rapidly evolved from a communication tool into a central repository for increasingly sensitive personal and financial data. Beyond simple contact lists, these devices now routinely store banking credentials, health records, government identification, and biometric information. This convergence of essential services within a single, portable device streamlines daily life, but simultaneously creates a powerful honeypot for malicious actors. The sheer volume of valuable data concentrated on smartphones, coupled with their ubiquitous presence, presents an unprecedented challenge to digital security. Individuals are now deeply reliant on these devices for everything from managing finances and accessing healthcare to proving identity and navigating daily routines, making the protection of smartphone data a critical concern for both individuals and institutions.

The approaching launch of the European Digital Identity (EUDI) Wallet dramatically underscores the increasing centrality of smartphones in managing sensitive personal data. This digital wallet, intended to provide secure and verifiable digital identification across Europe, will rely heavily on the security infrastructure of mobile devices to store and process credentials. Consequently, the EUDI initiative isn’t simply a software rollout; it’s a catalyst demanding a significant elevation of mobile security standards. The wallet’s functionality – encompassing everything from financial transactions to accessing public services – concentrates a vast array of sensitive information onto a single device, intensifying the need for robust cryptographic protections, secure storage mechanisms, and proactive threat detection. The EUDI Wallet, therefore, serves as a potent example of how smartphones are evolving into secure repositories of identity, and the implications for cybersecurity are substantial.

The pervasive integration of smartphones into daily life has inadvertently created a substantially expanded attack surface for malicious actors. These devices now function as central repositories for immensely sensitive data – from banking credentials and healthcare records to personally identifiable information – making them exceptionally valuable targets. This isn’t simply a matter of device theft; increasingly sophisticated cyberattacks target vulnerabilities within the smartphone’s operating system and applications, potentially enabling remote data extraction, identity theft, and financial fraud. The sheer number of users, combined with the complexity of modern mobile software and the constant introduction of new applications, presents a challenging landscape for security professionals and underscores the critical need for proactive threat mitigation and robust device protection strategies. Consequently, a compromised smartphone represents a far greater risk than ever before, extending beyond individual inconvenience to potentially widespread systemic vulnerabilities.

The security of modern smartphones increasingly relies on the strength of their cryptographic operations, yet these systems are not invulnerable. While designed to protect sensitive data like financial credentials and personal identification, cryptographic implementations on mobile devices present a complex attack surface. Researchers are actively investigating potential weaknesses in these systems, including side-channel attacks that exploit physical leakages during computation – such as power consumption or electromagnetic radiation – to deduce cryptographic keys. Furthermore, software vulnerabilities within the cryptographic libraries themselves, or improper implementation of cryptographic protocols, can expose devices to compromise. The advent of technologies like the EUDI Wallet, which will store highly sensitive digital identities on these same devices, underscores the urgent need for rigorous security assessments and the development of more robust cryptographic solutions tailored to the unique constraints and threats faced by mobile platforms.

Unseen Emanations: The Physics of Compromise

Electromagnetic (EM) Side-Channel Analysis is a decryption technique that exploits the unintentional electromagnetic radiation produced by electronic devices during cryptographic computations. This radiation, a byproduct of electrical current flow and switching activity, correlates directly with the data being processed and the operations being performed. By precisely measuring and analyzing these emanations with specialized equipment, attackers can deduce information about the cryptographic key without directly accessing the device’s memory or code. The analysis often involves statistical methods to filter noise and identify patterns related to key bits, allowing for key recovery even with relatively simple attacks. Unlike brute-force methods, EM Side-Channel Analysis can often recover keys in a single measurement, making it a highly efficient attack vector.

The susceptibility of cryptographic operations like the Conditional Swap within Elliptic Curve Digital Signature Algorithm (ECDSA) to Electromagnetic (EM) Side-Channel Analysis stems from data-dependent execution paths. These paths cause variations in power consumption and, consequently, electromagnetic emissions that correlate with processed data. Specifically, conditional branches – such as those within a conditional swap – execute different instructions based on key-dependent bits. This creates a discernible difference in the electromagnetic radiation emitted during each branch, allowing an attacker to infer information about the secret key by analyzing these emissions. Operations with highly data-dependent branching are thus significantly easier to exploit than those with consistent execution flows, as the signal-to-noise ratio is improved, increasing the accuracy of key recovery.

Electromagnetic (EM) Side-Channel Analysis circumvents conventional software security protocols such as firewalls, intrusion detection systems, and encryption algorithms by focusing on the physical characteristics of the cryptographic process itself. Rather than attempting to break the encryption mathematically or exploit software vulnerabilities, this technique analyzes the electromagnetic radiation unintentionally emitted by the hardware during computation. These emissions are directly correlated with the data being processed and the operations being performed, allowing an attacker to reconstruct the cryptographic key from the hardware’s physical output, independent of the software’s security features. This direct hardware targeting renders software-based defenses ineffective as the attack occurs at a lower level than the software execution environment.

The demonstrated effectiveness of Electromagnetic (EM) Side-Channel Analysis underscores the limitations of a security model predicated solely on software implementations. Software-based security measures, such as encryption algorithms and secure coding practices, operate within a defined computational space and are vulnerable when attackers bypass this layer. EM Side-Channel attacks exploit the physical characteristics of hardware during cryptographic operations-specifically, electromagnetic emanations-to derive sensitive information like cryptographic keys. This direct hardware targeting circumvents software protections entirely, revealing that robust security requires a holistic approach incorporating hardware-level countermeasures and physical security considerations in addition to software safeguards. Reliance on software alone creates a single point of failure susceptible to attacks that operate outside the scope of traditional software vulnerability assessments.

Implementing a 787 MHz bandpass filter significantly improves signal recovery within an Android application, as demonstrated by successful swap alignment in grey traces, though an unidentified interference source continues to hinder recovery in orange traces.

A Controlled Environment: The Fairphone 4 as a Research Platform

The Fairphone 4, equipped with the Qualcomm Snapdragon 750G 5G system-on-chip and running the PostmarketOS Linux distribution, provides a uniquely accessible platform for electromagnetic (EM) side-channel attack research on contemporary smartphones. Its modular design and open-source software stack allow for low-level access and customization, facilitating detailed analysis of the device’s cryptographic implementations. Unlike many modern smartphones with locked-down bootloaders and proprietary software, the Fairphone 4 enables researchers to directly observe and measure the electromagnetic emanations produced during cryptographic operations, which is essential for conducting EM side-channel analyses and evaluating the effectiveness of potential countermeasures. This openness positions the Fairphone 4 as a valuable tool for both academic investigation and security auditing of mobile device hardware and software.

The Fairphone 4’s open-source hardware and software architecture facilitates in-depth security analysis unavailable on most smartphones. Researchers benefit from access to schematics, board support packages, and the ability to modify the operating system, including kernel parameters and bootloaders. This level of control allows for precise measurements of electromagnetic emanations during cryptographic operations, enabling detailed side-channel analysis. Furthermore, the customizable software environment permits researchers to isolate and target specific hardware components and cryptographic implementations for vulnerability assessment, bypassing restrictions commonly imposed by proprietary systems.

Electromagnetic (EM) side-channel attacks were successfully demonstrated on the Fairphone 4, resulting in the recovery of cryptographic key bits. Through analysis of electromagnetic emanations during cryptographic operations, researchers were able to extract information related to the secret key. Specifically, the attacks yielded between 387 and 416 key bits per captured trace. Application of lattice recombination techniques suggests the potential for complete key recovery from these extracted bits. A custom Convolutional Neural Network (CNN) was employed for bit classification, achieving an accuracy of 98.14%, which slightly surpassed the performance of ASCAD networks in similar analyses. The attacks were performed with the device’s A77 processor operating at 787 MHz, and the CNN achieved an average probability of 90% for correct bit predictions, validating the feasibility of this attack vector on the Fairphone 4.

Electromagnetic side-channel analysis of the Fairphone 4 yielded recovery of 387 to 416 key bits per trace, indicating a significant vulnerability to this type of attack. Lattice recombination techniques offer a pathway to complete key recovery given the observed data rate. A custom Convolutional Neural Network (CNN) achieved a classifier accuracy of 98.14% in identifying key bits, demonstrating a slight performance improvement over established ASCAD networks. This CNN consistently predicted correct key bits with an average probability of 90%.

Electromagnetic side-channel attacks were successfully executed on the Fairphone 4’s A77 processor operating at a core frequency of 787 MHz. A custom Convolutional Neural Network (CNN) was employed to analyze the electromagnetic emanations, achieving an average probability of 90% for correct bit predictions. This indicates a high degree of correlation between the observed electromagnetic signals and the actual key bits being processed by the CPU, demonstrating the practical vulnerability of the device to this type of attack at the tested frequency. The consistent high probability of correct prediction suggests that the CNN effectively learned the relationship between the electromagnetic emissions and the internal cryptographic operations.

Photon emission measurements successfully identify active CPUs within the Snapdragon 750G 5G SoC by tracking activity during a simple benchmark program, differentiating between the six A55 CPUs (orange) and two A77 CPUs (blue).

The System’s Resilience: Countermeasures and the Pursuit of True Security

Protecting smartphones from Electromagnetic (EM) Side-Channel Attacks necessitates the implementation of robust countermeasures spanning both hardware and software domains. Hardware solutions often involve shielding sensitive components, adding noise to mask emissions, and employing balanced layouts to minimize radiation, while software techniques include randomization of execution, insertion of dummy operations, and constant-time algorithms. These strategies aim to disrupt the correlation between power consumption or electromagnetic emanations and the processed data, thereby hindering attackers’ ability to extract cryptographic keys or sensitive information. A layered approach, combining multiple countermeasures, is generally considered most effective, as it increases the complexity and cost of a successful attack. Ultimately, the efficacy of these defenses hinges on rigorous testing and continuous adaptation to increasingly sophisticated attack methodologies.

A cornerstone of modern smartphone security lies in the implementation of Secure Elements (SEs), specialized chips designed to isolate and protect sensitive cryptographic keys from compromise. These tamper-resistant components create a physically isolated environment, shielding keys used for encryption, authentication, and secure transactions from software-based attacks and electromagnetic emanations. By encapsulating critical data within a hardened boundary, the SE significantly reduces the attack surface available to malicious actors attempting to extract or manipulate confidential information. This isolation is particularly crucial given the increasing sophistication of side-channel attacks, where attackers exploit unintended physical leaks – such as electromagnetic radiation – to deduce secret keys. The utilization of SEs, therefore, represents a proactive and essential layer of defense, ensuring the integrity and confidentiality of user data and financial transactions on mobile devices.

Rigorous evaluation of any implemented countermeasure is essential, as theoretical security doesn’t guarantee practical resilience against Electromagnetic (EM) Side-Channel Analysis. Researchers must actively probe systems, employing sophisticated EM analysis techniques to determine the effectiveness of shields, noise injection, or algorithmic modifications intended to thwart attackers. This process involves meticulously measuring EM emanations during cryptographic operations, seeking vulnerabilities that could allow the recovery of secret keys – even with countermeasures in place. A successful defense isn’t simply about implementing security features, but demonstrably proving their ability to withstand real-world attacks, requiring a continuous cycle of testing, refinement, and validation to maintain a robust security posture.

The relentless evolution of electromagnetic (EM) side-channel attack techniques necessitates a continuous pursuit of innovative defensive strategies. Current countermeasures, while effective against known attacks, may prove insufficient against future, more sophisticated methods leveraging advancements in signal processing and machine learning. Ongoing research focuses on techniques like masking, where sensitive data is obscured with random values, and shuffling, which randomizes the order of operations to disrupt timing analysis. Furthermore, exploration into physically unclonable functions (PUFs) and truly random number generators promises to enhance key protection and cryptographic agility. This proactive approach is not merely about patching vulnerabilities; it’s about building inherently resilient systems capable of adapting to the ever-changing threat landscape and safeguarding the sensitive data entrusted to modern smartphones.

Analysis of OpenSSL/libgcrypt implementations with secp521r1, Ed25519, and secp128r1, using the [alam_nonce_at_once] countermeasure, reveals t-values and EM traces within defined trigger windows to assess potential leakage.

The pursuit of cryptographic security on modern smartphones, as demonstrated by this research into ECDSA vulnerabilities, reveals a fundamental truth about complex systems. Each layer of hardware and software, intended to fortify defenses, introduces new surfaces for attack. It’s a prophecy of eventual compromise, elegantly stated by G. H. Hardy: ‘The only way to do justice to the past is to make the present as different as possible.’ This study showcases how the ‘present’ of smartphone security, while vastly more intricate than previous iterations, still succumbs to the inherent limitations of physical implementation. The research highlights that even with secure enclaves, the electromagnetic emanations betray the underlying computations, reminding one that order – in this case, the illusion of cryptographic safety – is merely a temporary cache between inevitable failures.

What Lies Ahead?

The persistence of vulnerabilities to electromagnetic side-channel analysis, even within increasingly complex smartphone architectures, isn’t a failure of technique – it’s a predictable outcome. Each layer of abstraction, each secure enclave, merely shifts the surface area for attack, creating new resonant frequencies and propagation paths. The belief that software alone can solve hardware problems remains a comforting fiction. The work presented here demonstrates that the ecosystem adapts, finding weaknesses in the very protections intended to contain it.

Future research won’t focus on eliminating signal leakage – that’s akin to attempting to silence a thunderstorm. Instead, the field will likely gravitate toward more sophisticated noise models, adaptive attacks that exploit device-specific characteristics, and perhaps, a reluctant acceptance of inherent uncertainty. A guarantee of absolute security is simply a contract with probability; the cost of reducing that probability to zero is, ultimately, infinite.

Stability, as often perceived, is merely an illusion that caches well. The long-term trajectory suggests a constant arms race, not toward elimination of risk, but toward more subtle and resilient cryptographic implementations. The real challenge isn’t building unbreakable systems, but cultivating the ability to anticipate – and even accommodate – inevitable compromise. Chaos isn’t failure – it’s nature’s syntax.

Original article: https://arxiv.org/pdf/2512.07292.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

The Expanding Attack Surface: Smartphones as Data Concentrators

Unseen Emanations: The Physics of Compromise

A Controlled Environment: The Fairphone 4 as a Research Platform

The System’s Resilience: Countermeasures and the Pursuit of True Security

What Lies Ahead?

See also: