The bias–variance decomposition in profiled attacks

Lerman, Liran; Bontempi, Gianluca; Markowitch, Olivier

doi:10.1007/s13389-015-0106-1

Cited by 14 publications

(9 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our experiments essentially conclude that template attack outperform machine learning-based attacks whenever the number of dimensions can be kept reasonably low, e.g., thanks to a selection of points of interests (POI), and that machine learning (and random forest in particular) become(s) interesting in "extreme" profiling conditions (i.e., with large traces and a small profiling set)-which possibly arises when little information about the target device is available to the adversary. We then complement these results with an additional analysis based on the bias-variance decomposition of the error rate, which was recently introduced in the sidechannel literature [25]. The bias-variance decomposition allows separating the error rate of an attack in three weighted terms, among them the bias and the variance terms.…”

Section: Introductionmentioning

confidence: 90%

“…The goal of this section is to understand more deeply (i) why template attack can have a higher success rate than machine learning-based attack in a low dimensionality context, and (i) why a random forest outperforms template attack in a high dimensionality context. Our analyzes are based on the bias-variance decomposition of the error rate first proposed by Domingos in the field of machine learning [10,11] and then introduced in the side-channel literature by Lerman et al [25].…”

Section: Bias-variance Decomposition Analysismentioning

confidence: 99%

“…Recently, Lerman et al [25] showed that template attacks have a high variance while stochastic attack correspond to a trade-off between the bias and the variance terms. In this section, we aim to evaluate when and why template attacks generally worked well in our previous experiments, and machine learning algorithm (and more precisely random forests) can outperform them in extreme profiling conditions.…”

Section: Template Attackmentioning

confidence: 99%

See 2 more Smart Citations

Template attacks versus machine learning revisited and the curse of dimensionality in side-channel analysis: extended version

et al. 2017

Self Cite

View full text Add to dashboard Cite

Template attacks and machine learning are two popular approaches to profiled side-channel analysis. In this paper, we aim to contribute to the understanding of their respective strengths and weaknesses, with a particular focus on their curse of dimensionality. For this purpose, we take advantage of a well-controlled simulated experimental setting in order to put forward two important aspects. First and from a theoretic point of view, the data complexity of template attacks is not sensitive to the dimension increase in side-channel traces given that their profiling is perfect. Second and from a practical point of view, concrete attacks are always affected by (estimation and assumption) errors during profiling. As these errors increase, machine learning gains interest compared to template attacks, especially when based on random forests. We then clarify these results thanks to the bias-variance decomposition of the error rate recently introduced in the context side-channel analysis.

show abstract

Section: Introductionmentioning

confidence: 90%

Section: Bias-variance Decomposition Analysismentioning

confidence: 99%

Section: Template Attackmentioning

confidence: 99%

See 1 more Smart Citation

Template attacks versus machine learning revisited and the curse of dimensionality in side-channel analysis: extended version

et al. 2017

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the proposition can easily be generalised to other norms without impacting the conclusion. More precisely, Lerman et al show that three terms impact the error rate of profiled attacks: the noise, the bias (also known as the assumption error for TA) and the variance (also known as the estimation error for TA) [18]. The previous section analyses the bias term while this section compares the variance term of profiled attacks with simulations.…”

Section: B Estimation Errormentioning

confidence: 99%

“…Similarly to our work in [20], we focus on analytical and practical results obtained on simulated measurements on a software implementation. We base our theoretical results on the bias-variance theory (presented by Lerman et al [18] in side-channel attacks) and on the estimation/assumption errors (introduced by Durvaux et al [10]) in order to discover what impacts the success probability to retrieve the secret key. Furthermore, we report the performances of profiled attacks against an 8-bit Atmel XMEGA target device using using three kinds of actual measurements: (i) actual leakages with a low noise level, (ii) actual leakages with added noise and (iii) misaligned actual leakages.…”

Section: Introductionmentioning

confidence: 99%

Efficient Profiled Attacks on Masking Schemes

Lerman

Markowitch

2019

IEEE Trans.Inform.Forensic Secur.

Self Cite

View full text Add to dashboard Cite

Side-channel adversaries represent real-world threats against (certified and uncertified) cryptographic devices. Masking schemes represent prevailing countermeasures to reduce the success probabilities of side-channel attacks. However, masking schemes increase the implementation cost in term of power consumption, clock cycles, and random number generation. Investigation of tools evaluating the degree of resilience of cryptographic devices using masking (against side-channel attacks) represents an important aspect in certification procedures (e.g., Common Criteria, FIPS 140-2 and EMVco). Several side-channel evaluation techniques exist such as template attacks and machine learning based attacks. In this paper, we formalise results obtained in side-channel attacks community when targeting masked implementations. We report theoretical as well as practical results of parametric and non-parametric side-channel attacks on masking schemes. The theoretical part reports results based on a simulation of the execution of software devices while the practical part focuses on actual leakages measured during the execution of a software implementation in three different contexts (that contain different levels of noise).

show abstract

Climbing Down the Hierarchy: Hierarchical Classification for Machine Learning Side-Channel Attacks

Picek

Heuser

Jović

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

International audienceMachine learning techniques represent a powerful paradigm in side-channel analysis, but they come with a price. Selecting the appropriate algorithm as well as the parameters can sometimes be a difficult task. Nevertheless, the results obtained usually justify such an effort. However, a large part of those results use simplification of the data relation and in fact do not consider allthe available information. In this paper, we analyze the hierarchical relation between the data and propose a novel hierarchical classification approach for side-channel analysis. With this technique, we are able to introduce two new attacks for machine learning side-channel analysis: Hierarchical attack and Structured attack. Our results show that both attacks can outperform machine learning techniques using the traditional approach as well as the template attack regarding accuracy. To support our claims, we give extensive experimental results and discuss the necessary conditions to conduct such attacks

show abstract

The bias–variance decomposition in profiled attacks

Cited by 14 publications

References 34 publications

Template attacks versus machine learning revisited and the curse of dimensionality in side-channel analysis: extended version

Template attacks versus machine learning revisited and the curse of dimensionality in side-channel analysis: extended version

Efficient Profiled Attacks on Masking Schemes

Climbing Down the Hierarchy: Hierarchical Classification for Machine Learning Side-Channel Attacks

Contact Info

Product

Resources

About