Side-Channel Attacks: The Critical Vulnerability in Modern Cryptography
Executive Summary
Side-channel attacks represent a fundamental and often overlooked vulnerability in modern cryptographic systems. Unlike traditional cryptanalysis that targets mathematical weaknesses in encryption algorithms, side-channel attacks exploit physical implementation characteristics—electromagnetic emanations, power consumption, acoustic signatures, and timing variations—to extract cryptographic keys and sensitive data. This document provides a comprehensive examination of side-channel attack methodologies, documented real-world implementations, and their profound implications for cybersecurity. The evidence presented demonstrates that side-channel vulnerabilities constitute a significant gap in contemporary security architectures, capable of circumventing theoretically unbreakable encryption through signal intelligence and advanced signal processing techniques including Fourier transforms.
Introduction
Modern cryptographic algorithms such as Advanced Encryption Standard (AES) and Rivest-Shamir-Adleman (RSA) are mathematically robust, with properly implemented versions considered computationally infeasible to break through direct cryptanalysis[1]. However, the physical implementation of these algorithms in hardware and software creates unintended information leakage through various side channels. These physical emanations—whether electromagnetic radiation, power fluctuations, acoustic emissions, or timing variations—can be captured, analyzed using signal processing techniques, and correlated with secret cryptographic material[2].
The National Security Agency (NSA) has long recognized these vulnerabilities through programs such as TEMPEST, which focuses on preventing and exploiting electromagnetic emanations from electronic equipment[3]. The existence of such programs, combined with extensive academic research demonstrating practical attacks, establishes that side-channel vulnerabilities are not theoretical concerns but actively exploited security weaknesses.
Historical Context and Government Recognition
TEMPEST and Electromagnetic Emanations Security
The TEMPEST program, a code name referring to U.S. government research into compromising electromagnetic emanations, has existed since at least the 1960s[4]. TEMPEST encompasses the study of unintentional electromagnetic emanations from electronic devices that can leak classified information. The program's existence confirms governmental awareness that encryption alone provides insufficient protection against sophisticated adversaries capable of capturing and analyzing electromagnetic signals[5].
TEMPEST security measures include shielded rooms, specialized equipment certifications, and careful separation of systems processing classified information (Red systems) from those handling encrypted data (Black systems)[6]. The substantial investment in TEMPEST countermeasures by military and intelligence agencies demonstrates the practical reality and seriousness of electromagnetic side-channel threats.
Van Eck Phreaking
In 1985, Dutch researcher Wim van Eck publicly demonstrated that computer monitor displays could be reconstructed remotely by capturing electromagnetic emanations[7]. This technique, known as Van Eck phreaking, proved that specialized equipment could intercept the high-frequency electromagnetic radiation produced by cathode-ray tube (CRT) displays and reproduce the displayed content at a distance[8]. Modern flat-panel displays remain vulnerable through emanations from high-speed digital video interface (DVI) cables[9].
Van Eck's work demonstrated on BBC television that the technical capability existed to compromise visual information without any direct network or physical access to target systems[10]. This established electromagnetic eavesdropping as a practical attack vector rather than purely theoretical concern.
Side-Channel Attack Methodologies
Power Analysis Attacks
Power analysis attacks exploit correlations between electrical power consumption and data processing within cryptographic devices. These attacks are classified into two primary categories:
Simple Power Analysis (SPA)
Simple Power Analysis involves direct observation of power consumption traces to identify operations based on distinctive power signatures[11]. Attackers can distinguish between different cryptographic operations (such as multiplication versus squaring in RSA implementations) and identify conditional branches that depend on secret key material[12]. SPA attacks are relatively straightforward to execute but require implementation-specific knowledge.
Differential Power Analysis (DPA) and Correlation Power Analysis (CPA)
Differential Power Analysis and Correlation Power Analysis represent more sophisticated statistical techniques that correlate power consumption measurements with hypothetical intermediate computational values[13]. These attacks require no detailed knowledge of the implementation and can extract complete cryptographic keys from devices such as smart cards, embedded systems, and cryptographic accelerators[14].
Recent research has demonstrated successful Correlation Power Analysis attacks against AES implementations on FPGA platforms. One study showed that approximately 4,000 power traces enabled complete key extraction with 100% success rate using the Switching Distance power consumption model[15]. Another investigation successfully extracted cryptographic keys from ESP32 microcontrollers, breaking the hardware AES engine through power analysis despite the theoretical strength of the AES algorithm[16].
Research published in February 2026 documented advances in deep-learning-enhanced side-channel attacks on AES cryptography, demonstrating that machine learning models possess exceptional capacity to identify correlations between side-channel measurements and sensitive data, significantly enhancing attack effectiveness[2].
Electromagnetic Analysis Attacks
Electromagnetic (EM) analysis attacks capture and analyze electromagnetic radiation emitted during cryptographic operations. All electronic devices generate electromagnetic fields during operation, and these emanations contain information about the data being processed[17].
TEMPEST attacks specifically target these electromagnetic emissions. Recent documentation indicates that TEMPEST techniques can intercept electromagnetic signals from computers, monitors, keyboards, and wireless devices[18]. These signals can potentially reveal display information, keyboard activity patterns, and wireless communications[19].
Contemporary research continues to refine electromagnetic attack techniques. Modern devices with higher processing speeds and more complex components generate more pronounced electromagnetic emissions that sophisticated equipment can capture and analyze[20]. The attack requires proximity to the target device, with effectiveness influenced by distance, emission strength, and specialized reception equipment capabilities[21].
Acoustic Cryptanalysis
Acoustic side-channel attacks exploit sound emanations from computing hardware during cryptographic operations. In 2013, researchers from Tel Aviv University, including noted cryptographer Adi Shamir (co-inventor of RSA), demonstrated extraction of complete 4096-bit RSA decryption keys from laptop computers within one hour by analyzing acoustic signatures generated during decryption operations[22][23].
The attack methodology involved sending specially crafted ciphertexts that caused specific computational patterns during RSA decryption in GnuPG implementations. The researchers captured acoustic emanations using either a mobile phone placed adjacent to the target computer or a more sensitive microphone positioned four meters away[24]. Through thousands of repetitions and acoustic frequency spectrum analysis, researchers identified key-dependent sound patterns that enabled complete key extraction[25].
The acoustic cryptanalysis research demonstrated that multiple cryptographic implementations beyond RSA—including ElGamal decryption and other public-key systems—exhibited similar vulnerabilities to acoustic side-channel attacks[26]. This establishes acoustic analysis as a viable attack vector across diverse cryptographic systems.
Recent reports from February 2026 indicate continued research into acoustic attacks, with security experts noting that skilled attackers can extract cryptographic keys from keyboard typing sounds and other acoustic emanations from computing devices[27].
Timing Attacks
Timing attacks analyze variations in the execution time of cryptographic operations to infer information about secret keys. These attacks exploit the fact that cryptographic implementations often execute different code paths or require different numbers of operations depending on the secret key value being processed[28].
Cache-based timing attacks represent a particularly powerful variant. Modern processors utilize cache memory hierarchies to improve performance, but cache access patterns constitute a shared state between processes. Attackers can measure timing differences between cache hits and cache misses to determine which memory addresses a cryptographic operation accessed, thereby leaking information about secret keys[29][30].
The 2005 Bernstein cache timing attack against AES demonstrated practical key extraction through analysis of cache behavior[31]. Subsequent research has refined these techniques, showing that timing side channels remain exploitable across diverse cryptographic implementations.
Spectre and Meltdown: Microarchitectural Side Channels
The January 2018 disclosure of Spectre and Meltdown vulnerabilities revealed fundamental microarchitectural side channels affecting virtually all modern processors[32][33]. These attacks exploit speculative execution—a performance optimization where processors predict and execute instructions before determining whether execution should actually occur[34].
Spectre attacks train the processor's branch prediction mechanism to speculatively execute incorrect code paths that access privileged memory. Although the processor eventually discards the incorrect speculation results, the speculative execution leaves observable traces in cache memory[35]. Attackers combine speculative execution with cache timing side channels to extract sensitive information from other processes, virtual machines, and even kernel memory[36].
Meltdown similarly exploits speculative execution to bypass memory protection mechanisms, enabling unprivileged processes to read arbitrary kernel memory through timing-based observation of cache state[37]. The vulnerability affected Intel processors extensively, with hardware-level fixes requiring significant architectural changes beginning with 8th-generation Intel CPUs[38].
Research indicates that Spectre is potentially more powerful than Meltdown because it can exploit victim processes with specific vulnerable code patterns, whereas Meltdown primarily targets kernel memory within the same process[39]. Both vulnerabilities demonstrate that microarchitectural optimizations can introduce severe side-channel leakage despite cryptographic algorithms themselves remaining mathematically secure.
Recent Developments and Advanced Attacks
TEE.Fail: Attacking Trusted Execution Environments
Research published in October 2025 introduced TEE.Fail, a sophisticated side-channel attack targeting Trusted Execution Environments (TEEs) in modern processors[1]. The attack affects Intel's Software Guard Extensions (SGX) and Trust Domain Extensions (TDX), as well as AMD's Secure Encrypted Virtualization with Secure Nested Paging (SEV-SNP) and Ciphertext Hiding features.
TEE.Fail employs a physical interposition device constructed from off-the-shelf electronic equipment costing under $1,000. This device intercepts and analyzes memory traffic on DDR5 server buses, enabling extraction of cryptographic keys even from fully updated machines in trusted status[1]. Significantly, the attack succeeded against systems with constant-time cryptographic implementations and hardware-level encryption features specifically designed to prevent side-channel attacks, demonstrating that these countermeasures prove insufficient against sophisticated physical bus interposition techniques[40].
The research successfully extracted private signing keys from OpenSSL's ECDSA implementation despite the code being fully constant-time and running on systems with AMD's Ciphertext Hiding enabled[41]. This finding indicates that even implementations following cryptographic engineering best practices remain vulnerable to advanced side-channel attacks exploiting physical access to memory buses.
Deep Learning Enhanced Side-Channel Attacks
The integration of deep learning techniques with traditional side-channel analysis has significantly enhanced attack effectiveness. Machine learning models, particularly deep neural networks, excel at identifying complex correlations between physical measurements and secret cryptographic material without requiring detailed knowledge of the implementation[42].
Recent research has demonstrated that well-trained deep learning models can extract keys with dramatically fewer measurement traces compared to classical statistical techniques[2]. Studies show that first-order Differential Power Analysis attacks enhanced with deep learning can succeed with as few as 200 power traces against unprotected implementations[43].
Deep learning approaches have proven effective across multiple side-channel modalities, including power analysis, electromagnetic analysis, and cache timing attacks. The technique's ability to automatically learn relevant features from raw measurement data reduces the expertise required to execute successful attacks and increases the range of vulnerable implementations[44].
Furthermore, machine learning models themselves have become targets of side-channel attacks. Research from North Carolina State University demonstrated that power-based side-channel attacks can extract the secret weights and architecture of neural networks implemented in hardware, enabling intellectual property theft and model cloning[45].
Signal Processing Techniques in Side-Channel Attacks
Fourier Transform Analysis
Fourier transform techniques play a crucial role in side-channel attack signal processing. The Discrete Fourier Transform (DFT) analyzes power consumption traces, electromagnetic emissions, and acoustic signatures to extract frequency-domain information not readily apparent in time-domain signals[46][47].
Researchers employ Fourier analysis to identify periodic patterns in side-channel measurements that correlate with cryptographic operations. For acoustic cryptanalysis, Fourier transform processing isolates key-dependent frequency components in sound emanations from processors performing RSA operations[48]. In electromagnetic analysis, Fourier techniques identify radiation frequencies corresponding to specific data values being processed[49].
The Short-Time Fourier Transform (STFT) provides time-localized frequency analysis, enabling attackers to track how spectral characteristics change throughout cryptographic computations[50]. This temporal-frequency analysis proves particularly valuable when different key-dependent operations occur at different times during algorithm execution.
Wavelet Transform Applications
Wavelet transforms offer advantages over Fourier techniques for analyzing non-stationary signals common in side-channel measurements. Unlike Fourier transforms that provide only global frequency content, wavelet analysis delivers simultaneous time and frequency localization[51].
Research has documented four primary applications of wavelet transforms in side-channel analysis: acquisition of traces, pattern detection, noise filtering, and secret-key recovery[52]. Wavelet-based preprocessing techniques improve signal-to-noise ratios in power consumption measurements, enabling key extraction with fewer traces.
One study demonstrated that wavelet-based preprocessing combined with Empirical Mode Decomposition (EMD) reduced the number of traces required for successful key recovery by 17.7% compared to traditional attack methods[53]. The technique adaptively decomposes collected traces into intrinsic mode functions based on their own characteristics, requiring no prior knowledge about the cryptographic implementation[54].
Advanced Signal Processing Methodologies
Beyond Fourier and wavelet techniques, side-channel researchers employ sophisticated signal processing to overcome implementation-specific countermeasures. Desynchronization countermeasures introduce temporal variations that scatter sensitive information across multiple time positions in power traces, complicating correlation analysis[55].
To counter desynchronization, attackers apply signal transformation methods including higher-order cumulants, continuous wavelet transforms, and dynamic time warping algorithms[56]. These techniques realign or transform signals to reconstruct coherent patterns from deliberately scattered information.
Machine learning approaches increasingly replace or augment traditional signal processing. Convolutional neural networks can automatically learn relevant features from raw, misaligned traces without requiring manual signal preprocessing or synchronization[57]. This capability makes attacks more robust against common countermeasures.
Implications for Cybersecurity
The Encryption Paradox
Side-channel vulnerabilities create a fundamental paradox in cryptographic security: encryption algorithms may be mathematically unbreakable while remaining practically vulnerable through implementation characteristics. Organizations and individuals relying on encryption face a security gap between theoretical algorithm strength and real-world implementation security[58].
The extensive documentation of successful side-channel attacks against diverse systems—from embedded microcontrollers to high-security Trusted Execution Environments—demonstrates that this vulnerability spans the entire computing ecosystem[59]. No category of device appears immune, with successful attacks demonstrated against smart cards, mobile devices, servers, cryptographic accelerators, and even specially hardened secure processors[60].
Attack Accessibility and Cost
The decreasing cost and increasing accessibility of side-channel attack capabilities broadens the threat landscape. The TEE.Fail attack demonstrated that sophisticated physical side-channel attacks can be executed with equipment costing under $1,000[1]. Acoustic cryptanalysis requires only commodity mobile phones[61]. Software-based attacks like Spectre and Meltdown require no specialized hardware whatsoever[62].
This accessibility means that side-channel attacks no longer constitute threats exclusively from nation-state actors with substantial resources. Criminal organizations, corporate espionage operations, and even technically sophisticated individual attackers can potentially execute these attacks against high-value targets[63].
Countermeasure Limitations
While various countermeasures exist, the research evidence demonstrates significant limitations in their effectiveness. Constant-time implementations prevent timing attacks but remain vulnerable to power analysis, electromagnetic analysis, and physical interposition attacks[64]. Hardware-level encryption features like AMD's Ciphertext Hiding have been defeated by sophisticated attacks[65].
Electromagnetic shielding provides protection against remote eavesdropping but requires substantial investment and cannot be applied to portable devices or systems requiring normal user interaction[66]. Software-only countermeasures such as blinding, masking, and noise injection add computational overhead and complexity while providing incomplete protection[67].
The fundamental challenge is that side-channel countermeasures address specific attack vectors rather than eliminating the underlying information leakage. As researchers develop increasingly sophisticated signal processing and machine learning techniques, previously effective countermeasures become inadequate[68].
Impact on Post-Quantum Cryptography
The transition to post-quantum cryptography introduces additional side-channel concerns. While post-quantum algorithms resist attacks from quantum computers, their implementations remain vulnerable to classical side-channel attacks[69]. Recent evaluations of post-quantum cryptographic hardware implementations emphasize that side-channel security must be thoroughly assessed during the standardization and deployment process[70].
Post-quantum algorithms often involve more complex mathematical operations and larger key sizes compared to current cryptographic systems, potentially creating additional opportunities for side-channel leakage. Research specifically targeting side-channel attacks on post-quantum implementations has already begun, with documented successful attacks against lattice-based and code-based cryptographic schemes[71].
Medical Device and IoT Security Implications
The proliferation of Internet of Things (IoT) devices and AI-enabled medical devices creates particularly concerning attack surfaces for side-channel exploitation. IoT devices typically employ constrained hardware with limited resources for implementing comprehensive side-channel countermeasures[72].
Medical devices incorporating cryptographic protections for patient data confidentiality and device integrity remain vulnerable to physical and side-channel attacks[73]. The FDA's cybersecurity requirements for AI medical devices acknowledge that dedicated attackers could physically tamper with device hardware or employ side-channel techniques, though such attacks are considered less common than network-based threats[74].
The embedded nature of IoT and medical devices often provides attackers with physical proximity, eliminating distance-based limitations on electromagnetic and acoustic side-channel attacks. Furthermore, the long operational lifetimes of embedded systems mean that devices deployed today may face increasingly sophisticated attacks over decades of service without feasible hardware security updates[75].
Conclusion
Side-channel attacks represent a critical and persistent vulnerability in modern cryptographic systems. The extensive body of research—spanning government programs like TEMPEST, academic demonstrations of acoustic cryptanalysis, and sophisticated attacks on cutting-edge Trusted Execution Environments—establishes conclusively that encryption alone provides insufficient security against sophisticated adversaries.
The fundamental challenge lies in the gap between cryptographic theory and physical implementation. Mathematically unbreakable algorithms leak sensitive information through electromagnetic emanations, power consumption patterns, acoustic signatures, timing variations, and microarchitectural state changes. Advanced signal processing techniques, including Fourier transforms, wavelet analysis, and machine learning, enable extraction of cryptographic keys and sensitive data from these physical side channels.
The accessibility of side-channel attack capabilities continues to increase while costs decrease, broadening the threat landscape beyond nation-state actors. Countermeasures remain incomplete and often inadequate against sophisticated attacks combining multiple techniques. Organizations relying on cryptographic protections must recognize that side-channel vulnerabilities constitute a substantial gap in security architectures—a gap that determined adversaries with appropriate expertise and equipment can exploit to completely circumvent even theoretically unbreakable encryption.
Future security architectures must address side-channel vulnerabilities through comprehensive defense-in-depth approaches combining physical security, electromagnetic shielding, algorithmic countermeasures, secure hardware design, and recognition that complete elimination of information leakage may be fundamentally impossible given the physical nature of computation. The evidence presented establishes that side-channel attacks are not merely theoretical concerns but practical, documented, and actively exploited vulnerabilities that significantly compromise the security posture of modern cryptographic systems.
References
[1] The Hacker News. (2025, October 27). New TEE.Fail Side-Channel Attack Extracts Secrets from Intel and AMD Processors. https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html
[2] Wang, X., et al. (2026). Recent Advances in Deep-Learning Side-Channel Attacks on AES Cryptography. Computers, Materials & Continua, 87(1). https://www.techscience.com/cmc/v87n1/66095
[3] Panda Security. (2024, January 16). TEMPEST and EMSEC: Is it possible to use electromagnetic waves to steal data? https://www.pandasecurity.com/en/mediacenter/cyberattack-electromagnetic-emanations/
[4] Kuhn, M. (1998). Hidden Data Transmission Using Electromagnetic Emanations. University of Cambridge Computer Laboratory. https://www.cl.cam.ac.uk/~mgk25/ih98-tempest.pdf
[5] Grey Dynamics. (2025, November 28). TEMPEST: Electronic Spying and Countermeasures. https://greydynamics.com/tempest-electronic-spying-and-countermeasures/
[6] Panda Security. (2024, January 16). TEMPEST protection measures and Red/Black architecture. https://www.pandasecurity.com/en/mediacenter/cyberattack-electromagnetic-emanations/
[7] van Eck, W. (1985). Electromagnetic Radiation from Video Display Units: An Eavesdropping Risk? Computers & Security, 4(4), 269-286.
[8] Startup Defense. (2026, February 5). TEMPEST Attacks: A Technical Guide to Digital Security. https://www.startupdefense.io/cyberattacks/tempest-attack
[9] Wikipedia. (2004). Van Eck phreaking. https://en.wikipedia.org/wiki/Van_Eck_phreaking
[10] Startup Defense. (2026, February 5). BBC demonstration of TEMPEST capabilities. https://www.startupdefense.io/cyberattacks/tempest-attack
[11] Kocher, P., et al. (1999). Differential Power Analysis. Advances in Cryptology—CRYPTO '99, Lecture Notes in Computer Science, 1666, 388-397.
[12] Semantic Scholar. (n.d.). Differential Power Analysis attacks on AES. https://pdfs.semanticscholar.org/e92d/50e613fc38849cf39940eaeea914869c6c51.pdf
[13] CAD4Security. (n.d.). Chapter 8: Power Side-Channel Analysis on AES. http://cad4security.org/index.php/trainings/hsl/ch8_psc_on_aes/
[14] Raelize. (2023, February 23). Espressif ESP32: Breaking HW AES with Power Analysis. https://raelize.com/blog/espressif-systems-esp32-breaking-hw-aes-with-power-analysis/
[15] Ali, S., et al. (2025, February 1). Evaluating AES Security: Correlation Power Analysis Attack. Engineering, Technology & Applied Science Research. https://etasr.com/index.php/ETASR/article/view/9728
[16] Raelize. (2023, February 23). Breaking ESP32 hardware AES engine using power analysis. https://raelize.com/blog/espressif-systems-esp32-breaking-hw-aes-with-power-analysis/
[17] Semiconductor Engineering. (2025). Side Channel Attacks. https://semiengineering.com/knowledge_centers/semiconductor-security/side-channel-attacks/
[18] Startup Defense. (2026, February 5). TEMPEST attack electromagnetic interception techniques. https://www.startupdefense.io/cyberattacks/tempest-attack
[19] Startup Defense. (2026, February 5). Information vulnerable to TEMPEST attacks. https://www.startupdefense.io/cyberattacks/tempest-attack
[20] Startup Defense. (2026, February 5). Modern device electromagnetic emissions. https://www.startupdefense.io/cyberattacks/tempest-attack
[21] Startup Defense. (2026, February 5). Factors affecting TEMPEST attack effectiveness. https://www.startupdefense.io/cyberattacks/tempest-attack
[22] Genkin, D., Shamir, A., & Tromer, E. (2013). RSA Key Extraction via Low-Bandwidth Acoustic Cryptanalysis. https://cs-people.bu.edu/tromer/acoustic/
[23] Schneier, B. (2020, January 26). Acoustic Cryptanalysis. Schneier on Security. https://www.schneier.com/blog/archives/2013/12/acoustic_crypta.html
[24] Linux Security. (2013, December 18). Acoustic Cryptanalysis Attack: GnuPG RSA Key Extraction. https://linuxsecurity.com/news/cryptography/acoustic-cryptanalysis
[25] Security Affairs. (2013, December 20). Breaking 4096-bit RSA with an acoustic Cryptanalysis attack. https://securityaffairs.com/20637/hacking/acoustic-cryptanalysis-attack.html
[26] Security Affairs. (2013, December 20). Acoustic cryptanalysis of ElGamal and other systems. https://securityaffairs.com/20637/hacking/acoustic-cryptanalysis-attack.html
[27] Portnox. (2026, February 8). Your Keyboard Might Be Betraying You: Acoustic Attacks. https://www.portnox.com/blog/endpoint-security/why-your-keyboard-might-be-betraying-you-acoustic-attacks-the-secure-alternatives
[28] Twingate. (2025, October 6). What is a Timing Attack? How It Works & Examples. https://www.twingate.com/blog/glossary/timing%20attack
[29] Hu, G. (2020, June 9). CPU Cache Side-Channel Attacks: Meltdown & Spectre. https://www.josehu.com/technical/2020/06/10/cpu-side-channel.html
[30] Xiao, D., & Mihretie, Y. (2022). Timing Attacks on Cryptographic Algorithms. MIT. https://courses.csail.mit.edu/6.857/2022/projects/Xiao-Mihretie.pdf
[31] Wienecke, M. (2021). Cache based Timing Attacks on Embedded Systems. Ruhr University Bochum. https://informatik.rub.de/wp-content/uploads/2021/11/ms_wienecke.pdf
[32] Meltdown Attack. (2018, January 14). Meltdown and Spectre. https://meltdownattack.com
[33] Wikipedia. (2018). Spectre (security vulnerability). https://en.wikipedia.org/wiki/Spectre_(security_vulnerability)
[34] Arm Developer. (2025, July 22). Arm CPU Security Bulletin: Spectre/Meltdown. https://developer.arm.com/documentation/110280/latest/
[35] Hu, G. (2020, June 9). Spectre attack mechanism and cache timing channels. https://www.josehu.com/technical/2020/06/10/cpu-side-channel.html
[36] Xiao, D., & Mihretie, Y. (2022). Spectre attacks on cryptographic systems. MIT. https://courses.csail.mit.edu/6.857/2022/projects/Xiao-Mihretie.pdf
[37] Hu, G. (2020, June 9). Meltdown kernel memory access vulnerability. https://www.josehu.com/technical/2020/06/10/cpu-side-channel.html
[38] Reddit. (2025, July 13). Windows 11 - Spectre/Meltdown Side Channel Attacks. https://www.reddit.com/r/Windows11/comments/1lz3v8l/windows_11_spectremeltdown_side_channel_attacks/
[39] Hu, G. (2020, June 9). Comparing Spectre and Meltdown capabilities. https://www.josehu.com/technical/2020/06/10/cpu-side-channel.html
[40] The Hacker News. (2025, October 27). TEE.Fail attack against constant-time implementations. https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html
[41] The Hacker News. (2025, October 27). Extraction of OpenSSL ECDSA keys despite countermeasures. https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html
[42] Bursztein, E. (2021, July 14). Hacker's guide to deep-learning side-channel attacks: the theory. https://elie.net/blog/security/hacker-guide-to-deep-learning-side-channel-attacks-the-theory
[43] Rambus. (2020, June 16). Side-Channel Attacks Target Machine Learning (ML) Models. https://www.rambus.com/blogs/side-channel-attacks-target-machine-learning-ml-models/
[44] Wang, X., et al. (2026). Deep learning model capabilities in side-channel analysis. Computers, Materials & Continua, 87(1). https://www.techscience.com/cmc/v87n1/66095
[45] Rambus. (2020, June 16). Power-based attacks on neural network implementations. https://www.rambus.com/blogs/side-channel-attacks-target-machine-learning-ml-models/
[46] NIST. (n.d.). Novel Applications of Wavelet Transforms based Side-Channel Analysis. https://csrc.nist.gov/csrc/media/events/non-invasive-attack-testing-workshop/documents/01_souissi.pdf
[47] Taylor & Francis. (2019, August 31). Side-channel attacks – Signal transformation methods. https://taylorandfrancis.com/knowledge/Engineering_and_technology/Computer_science/Side-channel_attacks/
[48] Security Affairs. (2013, December 20). Acoustic frequency spectrum analysis for key extraction. https://securityaffairs.com/20637/hacking/acoustic-cryptanalysis-attack.html
[49] Larksuite. (2024, May 28). Van Eck Phreaking electromagnetic analysis. https://www.larksuite.com/en_us/topics/cybersecurity-glossary/van-eck-phreaking
[50] NIST. (n.d.). Short-Time Fourier Transform applications in side-channel analysis. https://csrc.nist.gov/csrc/media/events/non-invasive-attack-testing-workshop/documents/01_souissi.pdf
[51] NIST. (n.d.). Wavelet transform advantages for side-channel analysis. https://csrc.nist.gov/csrc/media/events/non-invasive-attack-testing-workshop/documents/01_souissi.pdf
[52] NIST. (n.d.). Four applications of wavelet transforms in side-channel context. https://csrc.nist.gov/csrc/media/events/non-invasive-attack-testing-workshop/documents/01_souissi.pdf
[53] Wiley. (2019, October 29). A Highly Effective Data Preprocessing in Side-Channel Attack Using Empirical Mode Decomposition. https://onlinelibrary.wiley.com/doi/10.1155/2019/6124165
[54] Wiley. (2019, October 29). EMD adaptive decomposition methodology. https://onlinelibrary.wiley.com/doi/10.1155/2019/6124165
[55] Taylor & Francis. (2019, August 31). Misalignment countermeasures and signal transformation. https://taylorandfrancis.com/knowledge/Engineering_and_technology/Computer_science/Side-channel_attacks/
[56] Taylor & Francis. (2019, August 31). Signal transformation methods for overcoming countermeasures. https://taylorandfrancis.com/knowledge/Engineering_and_technology/Computer_science/Side-channel_attacks/
[57] Bursztein, E. (2021, July 14). Deep learning for automated feature extraction from raw traces. https://elie.net/blog/security/hacker-guide-to-deep-learning-side-channel-attacks-the-theory
[58] Bursztein, E. (2021, July 14). Side-channel attacks target implementation rather than algorithm. https://elie.net/blog/security/hacker-guide-to-deep-learning-side-channel-attacks-the-theory
[59] Wang, X., et al. (2026). Side-channel vulnerabilities across IoT ecosystem. Computers, Materials & Continua, 87(1). https://www.techscience.com/cmc/v87n1/66095
[60] The Hacker News. (2025, October 27). Side-channel attacks spanning device categories. https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html
[61] Genkin, D., Shamir, A., & Tromer, E. (2013). Mobile phone as acoustic cryptanalysis equipment. https://cs-people.bu.edu/tromer/acoustic/
[62] Wikipedia. (2018). Spectre and Meltdown software-based attacks. https://en.wikipedia.org/wiki/Spectre_(security_vulnerability)
[63] Bursztein, E. (2021, July 14). Democratization of side-channel attack capabilities. https://elie.net/blog/security/hacker-guide-to-deep-learning-side-channel-attacks-the-theory
[64] The Hacker News. (2025, October 27). Constant-time implementation limitations. https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html
[65] The Hacker News. (2025, October 27). AMD Ciphertext Hiding defeated by TEE.Fail. https://thehackernews.com/2025/10/new-teefail-side-channel-attack.html
[66] Grey Dynamics. (2025, November 28). TEMPEST countermeasures and limitations. https://greydynamics.com/tempest-electronic-spying-and-countermeasures/
[67] Wienecke, M. (2021). Software countermeasure effectiveness and overhead. Ruhr University Bochum. https://informatik.rub.de/wp-content/uploads/2021/11/ms_wienecke.pdf
[68] Wang, X., et al. (2026). Evolution of attack techniques defeating countermeasures. Computers, Materials & Continua, 87(1). https://www.techscience.com/cmc/v87n1/66095
[69] PQShield. (2025, November 25). Evaluating the Side Channel Security of Post-Quantum Hardware IP. https://pqshield.com/evaluating-the-side-channel-security-of-post-quantum-hardware-ip/
[70] PQShield. (2025, November 25). Side-channel evaluation requirements for post-quantum cryptography. https://pqshield.com/evaluating-the-side-channel-security-of-post-quantum-hardware-ip/
[71] Semiconductor Engineering. (2025, October 9). Side-Channel Attacks On Post-Quantum Cryptography. https://semiengineering.com/knowledge_centers/semiconductor-security/side-channel-attacks/
[72] Wang, X., et al. (2026). IoT device side-channel vulnerabilities and constraints. Computers, Materials & Continua, 87(1). https://www.techscience.com/cmc/v87n1/66095
[73] Intuition Labs. (2026, February 11). AI Medical Device Cybersecurity: Physical and Side-Channel Attacks. https://intuitionlabs.ai/articles/cybersecurity-requirements-ai-medical-devices
[74] Intuition Labs. (2026, February 11). FDA cybersecurity requirements acknowledging side-channel threats. https://intuitionlabs.ai/articles/cybersecurity-requirements-ai-medical-devices
[75] Wang, X., et al. (2026). Long-term vulnerabilities in embedded systems. Computers, Materials & Continua, 87(1). https://www.techscience.com/cmc/v87n1/66095