Idempotent Boolean functions form a highly structured subclass of Boolean functions that is closely related to rotation symmetry under a normal-basis representation and to invariance under a fixed linear map in a polynomial basis. These functions are attractive as candidates for cryptographic design, yet their additional algebraic constraints make the search for high nonlinearity substantially more difficult than in the unconstrained case. In this work, we investigate evolutionary methods for constructing highly nonlinear idempotent Boolean functions for dimensions n = 5 up to n = 12 using a polynomial basis representation with canonical primitive polynomials. Our results show that the problem of evolving idem-potent functions is difficult due to the disruptive nature of crossover and mutation operators. Next, we show that idempotence can be enforced by encoding the truth table on orbits, yielding a compact genome of size equal to the number of distinct squaring orbits.

Federated Transfer Learning (FTL) is the most general variation of Federated Learning. According to this distributed paradigm, a feature learning pre-step is commonly carried out by only one party, typically the server, on publicly shared data. After that, the Federated Learning phase takes place to train a classifier collaboratively using the learned feature extractor. Each involved client contributes by locally training only the classification layers on a private training set. The peculiarity of an FTL scenario makes it hard to understand whether poisoning attacks can be developed to craft an effective backdoor. State-of-the-art attack strategies assume the possibility of shifting the model attention toward relevant features introduced by a forged trigger injected in the input data by some untrusted clients. Of course, this is not feasible in FTL, as the learned features are fixed once the server performs the pre-training step. Consequently, in this paper, we investigate this intriguing Federated Learning scenario to identify and exploit a vulnerability obtained by combining eXplainable AI (XAI) and dataset distillation. In particular, the proposed attack can be carried out by one of the clients during the Federated Learning phase of FTL by identifying the optimal local for the trigger through XAI and encapsulating compressed information of the backdoor class. Due to its behavior, we refer to our approach as a focused backdoor approach (FB-FTL for short) and test its performance by explicitly referencing an image classification scenario. With an average 80% attack success rate, obtained results show the effectiveness of our attack also against existing defenses for Federated Learning.

While security vulnerabilities in traditional Deep Neural Networks have been extensively studied, the susceptibility of Spiking Neural Networks (SNNs) to adversarial attacks remains mostly underexplored. In fact, until now, the mechanisms for injecting backdoors into SNN models have been limited to digital scenarios. In this work, we present the first evaluation of backdoor attacks on SNN models in real-world physical environments, using event-based Dynamic Vision Sensor cameras. We assess and identify the limitations of existing backdoors in physical settings. To address each limitation, we develop three novel backdoor attack methods on SNNs, i.e., Framed, Strobing, and Flashy Backdoor, each progressively enhancing attack effectiveness and physical transferability. Our methods achieve up to a 100% Attack Success Rate with a negligible drop in clean accuracy across all tested datasets. We adapt and evaluate the effectiveness of state-of-the-art backdoor defenses from the image domain for SNNs. Next, we assess trigger stealthiness with commonly used metrics, finding them highly stealthy. Finally, we propose alternative detection techniques better suited for neuromorphic data. The code, new dataset, and results are available in our repository.11https://github.com/Yencr0s/Flashy_backdoor

Backdoor attacks maliciously inject covert functionality into machine learning models, which has been considered a security threat. The stealthiness of backdoor attacks is a critical research direction, focusing on adversaries' efforts to enhance the resistance of backdoor attacks against defense mechanisms. Recent research on backdoor stealthiness focuses mainly on indistinguishable triggers in input space and inseparable backdoor representations in feature space, aiming to circumvent backdoor defenses that examine these respective spaces. However, existing backdoor attacks are typically designed to resist a specific type of backdoor defense without considering the diverse range of defense mechanisms. Based on this observation, we pose a natural question: Are current backdoor attacks truly a real-world threat when facing diverse practical defenses?

Machine learning (ML) has been widely deployed in various applications, with many applications being in critical infrastructures. One recent paradigm is edge ML, an implementation of ML on embedded devices for Internet-of- Things (IoT) applications. In this work, we have conducted a practical experiment on Intel Neural Compute Stick (NCS) 2, an edge ML device, with regard to fault injection (FI) attacks. More precisely, we have employed electromagnetic fault injection (EMFI) on NCS 2 to evaluate the practicality of the attack on a real target device. We have investigated multiple fault parameters with a low-cost pulse generator, aiming to achieve misclassification at the output of the inference. Our experimental results demonstrated the possibility of achieving practical and repeatable misclassifications.

Large Language Models (LLMs) are susceptible to various attacks but can also improve the security of diverse systems. However, how well do open source LLMs behave as covertext distributions to, e.g., facilitate censorship-resistant communication? In this paper, we explore open-source LLM-based covert channels. We empirically measure the security vs. capacity of two open-source LLM models (Llama-7B and GPT-2) to assess their performance as covert channels. Although our results indicate that such channels are not likely to achieve high practical bitrates, we also show that the chance for an adversary to detect covert communication is low. To ensure our results can be used with the least effort as a general reference, we employ a conceptually simple and concise scheme and only assume public models.

Side-channel analysis (SCA) poses a real-world threat by exploiting unintentional physical signals to extract secret information from secure devices. Evaluation labs also use the same techniques to certify device security. In recent years, deep learning has emerged as a prominent method for SCA, achieving state-of-the-art attack performance at the cost of interpretability. Understanding how neural networks extract secrets is crucial for security evaluators aiming to defend against such attacks, as only by understanding the attack can one propose better countermeasures. In this work, we apply mechanistic interpretability to neural networks trained for SCA, revealing $\textit{how}$ models exploit $\textit{what}$ leakage in side-channel traces. We focus on sudden jumps in performance to reverse engineer learned representations, ultimately recovering secret masks and moving the evaluation process from black-box to white-box. Our results show that mechanistic interpretability can scale to realistic SCA settings, even when relevant inputs are sparse, model accuracies are low, and side-channel protections prevent standard input interventions.

Non-normal Boolean bent functions are one of the least understood classes of bent functions, and only a few difficult-to-find examples of such functions are known. In this paper, we consider the following two open problems on the normality of bent functions: 1. Do non-normal bent functions in 8 variables and degree 4 exist? 2. Do non-normal bent functions in the PS − ∖ PS a p class exist? We solve both of these problems by finding among the known PS bent functions in n = 8 variables a non-normal bent function in the PS − ∖ PS a p class.

Point of Sale (PoS) machines have become extremely popular recently. In many economies, most transactions occur using them. Although PoS technology is evolving, PINs are still heavily used. In this paper, we perform a large-scale study to understand how difficult it is to uncover user PINs at PoS, even when the users cover the pad with their hands. Our study involves 142 participants, two types of PoS, and around 13,800 PINs. We develop machine learning techniques to infer PoS PINs by using hidden cameras. Our results show that uncovering PINs in PoS is more complex than in other cases where a user PIN is used, e.g., ATMs, because of the small pad area of PoS. Nevertheless, we could achieve more than 50% Top-3 accuracy for 4-digit PINs and 45% Top-3 accuracy for 5-digit PINs, even when the PIN is covered by the user's hand. We comment on the impact of the camera's position and PoS on the successful inference of the user's PINs. We also comment on the hardness of inferring PINs depending on the physical distance of digits and recommend what are good practices to generate PINs and cover PoS to make PIN inference difficult.

Membership Inference Attacks (MIAs) infer whether a data point is in the training data of a machine learning model, posing privacy risks to sensitive data like medical records or financial data. Intuitively, data points that MIA accurately detects are vulnerable. Those data points may exist in the data of different target models, each susceptible to multiple MIAs. As such, the vulnerability of data points under multiple MIAs and target models represents a significant challenge. This article defines several metrics reflecting data points’ vulnerability and capturing vulnerable data points under multiple MIAs and target models. We implement 77 MIAs, with an average attack accuracy over target models ranging from 0.5 to 0.9, to support our analysis with our scalable and flexible platform, Various Membership Inference Attacks Platform (VMIAP). Based on the results, we observe that MIA has an inference tendency to some data points despite a low overall inference performance. Furthermore, previous approaches are unsuitable for finding vulnerable data points under multiple MIAs and target models. Finally, we explore the impact of retraining target, shadow, and attack models separately on the vulnerability of data points.

Side-channel attacks cause a significant threat to security implementations in embedded devices. Accordingly, an automated framework simulating side-channel behaviours can offer invaluable insights into leakage origins and characteristics, helping developers improve those devices during the design phase. While there has been a substantial effort towards crafting leakage simulators, earlier methods either necessitated significant manual work for reverse engineering the micro-architectural layer or depended on Deep Learning (DL) models where the neural network's complexity increased considerably with the addition of pipeline stages. This paper presents a novel modelling approach using Recurrent Neural Networks (RNNs) to construct instruction-level power models that exhibit enhanced performance in detecting pipeline leakage. Our findings indicate that with memory-based machine learning models, it becomes unnecessary to input data accounting for the pipeline effect. This strategy reduces feature dimensionality by at least one-third for a three-stage pipeline, albeit at a modest compromise in model performance. This reduced feature set underscores our model's scalability, making it a preferred choice for analyzing microprocessors with extended pipeline stages. Importantly, our methodology accelerates the micro-architectural profiling phase in side-channel simulator design. When evaluated on an expansive dataset, the performance of our memory-based model closely matches that of the Multilayer Perceptron (MLP) with an R2 value of 0.79. On a reduced dataset (removing the pipeline effect), our model achieves an R2 value of 0.65, outperforming the MLP, which reaches an R2 value of 0.39. Moreover, our model is designed with scalability in mind, making it suitable for profiling microcontrollers with advanced pipeline stages. For the practical realisation of our approach, we employed the open-source ABBY-CM0 dataset from the ARM Cortex-M0 microcontroller, which has three pipeline stages. To provide a detailed analysis, we also consider a Convolutional Neural Network (CNN) besides two RNN architectures (Long Short-Term Memory and Gated Recurrent Unit).

In this paper, we conduct a detailed power side-channel analysis of Xoodyak, a lightweight cryptographic algorithm, on an FPGA platform. We focus on the initialization phase of Xoodyak in the authenticated encryption with associated data (AEAD) mode. First, we introduce a new leakage model and perform a leakage assessment. Then, we perform non-profiled and profiled attacks to determine if the observed leakages can be exploited. For a non-profiled attack, we perform a correlation power analysis on all key bits, achieving a success rate of 91.4% with 50 000 traces. Our approach for a profiled attack involves a template attack and a deep learning-based attack. The former achieves a success rate of 99.2%, recovering almost all key bits with 20 000 traces in the attack phase. The latter reaches a guessing entropy of zero after 550 traces and adapts to the leakage model within 50 epochs.

Non-cryptographic (NC) hash functions are crucial in high-speed search applications and probabilistic data structures (PDS) such as Bloom filters and Count-Min sketches for efficient lookups and counting. These operations necessitate execution at line rates to accommodate the high-speed demands of Terabit Ethernet networks, characterized by bandwidths exceeding 100 Gbps. Consequently, a growing inclination towards hardware platforms, particularly Field Programmable Gate Arrays (FPGAs), is evident in network security applications. Given the centrality of hash functions in these structures, any enhancements to their design carry substantial implications for overall system performance. However, hash functions must exhibit independence, uniform distribution, and hardware-friendly characteristics. In this work, we employ Genetic Programming (GP) with avalanche metrics as a fitness function to devise a hardware-friendly family of NC hash functions called the Evolutionary hash (E-hash). We provide a detailed experimental analysis to offer insights on primitive set combinations involving logical operations and diverse hyperparameter settings, encompassing variables such as the number of nodes, tree height, population size, crossover and mutation rate, tournament size, number of constants, and generations. Compared to existing state-of-the-art hardware-friendly hash functions, the proposed E-hash family exhibits an 8.4% improvement in terms of operating frequency and throughput and 7.74% in latency on FPGA.

Backdoor attacks on deep learning represent a recent threat that has gained significant attention in the research community. Backdoor defenses are mainly based on backdoor inversion, which has been shown to be generic, model-agnostic, and applicable to practical threat scenarios. State-of-the-art backdoor inversion recovers a mask in the feature space to locate prominent backdoor features, where benign and backdoor features can be disentangled. However, it suffers from high computational overhead, and we also find that it overly relies on prominent backdoor features that are highly distinguishable from benign features. To tackle these shortcomings, this paper improves backdoor feature inversion for backdoor detection by incorporating extra neuron activation information. In particular, we adversarially increase the loss of backdoored models with respect to weights to activate the backdoor effect, based on which we can easily differentiate backdoored and clean models. Experimental results demonstrate our defense, BAN, is 1.37$\times$ (on CIFAR-10) and 5.11$\times$ (on ImageNet200) more efficient with an average 9.99\% higher detect success rate than the state-of-the-art defense BTI DBF. Our code and trained models are publicly available at https://github.com/xiaoyunxxy/ban.

The use of deep learning-based side-channel analysis is an effective way of performing profiling attacks on power and electromagnetic leakages, even against targets protected with countermeasures. While many research papers have reported successful results, they typically focus on profiling and attacking a single device, assuming that leakages are similar between devices of the same type. However, this assumption is not always realistic due to variations in hardware and measurement setups, creating what is known as the portability problem. Profiling multiple devices has been proposed as a solution, but obtaining access to these devices may pose a challenge for attackers. This paper proposes a new approach to overcome the portability problem by introducing a neural network layer assessment methodology based on the ablation paradigm. This methodology evaluates the sensitivity and resilience of each layer, providing valuable knowledge to create a Multiple Device Model from Single Device (MDMSD). Specifically, it involves ablating a specific neural network section and performing recovery training. As a result, the profiling model, trained initially on a single device, can be generalized to leakage traces measured from various devices. By addressing the portability problem through a single device, practical side-channel attacks could be more accessible and effective for attackers.

In side-channel analysis (SCA), Points-of-Interest (PoIs), i.e., the informative sample points remain sparsely scattered across the whole side-channel trace. Several works in the SCA literature have demonstrated that the attack efficacy could be significantly improved by combining information from the sparsely occurring PoIs. In Deep Learning (DL), a common approach for combining the information from the sparsely occurring PoIs is softmax attention. This work studies the training instability of the softmax attention-based CNN models on long traces. We show that the softmax attention-based CNN model incurs an unstable training problem when applied to longer traces (e.g., traces having a length greater than$10K$sample points). We also explore the use of batch normalization and multi-head softmax attention to make the CNN models stable. Our results show that the use of a large number of batch normalization layers and/or multi-head softmax attention (replacing the vanilla softmax attention) can make the models significantly more stable, resulting in better attack efficacy. Moreover, we found our models to achieve similar or better results (up to 85% reduction in the minimum number of the required traces to reach the guessing entropy 1) than the state-of-the-art results on several synchronized and desynchronized datasets. Finally, by plotting the loss surface of the DL models, we demonstrate that using multi-head softmax attention instead of vanilla softmax attention in the CNN models can make the loss surface significantly smoother.

Recently, researchers have successfully employed Graph Neural Networks (GNNs) to build enhanced recommender systems due to their capability to learn patterns from the interaction between involved entities. In addition, previous studies have investigated federated learning as the main solution to enable a native privacy-preserving mechanism for the construction of global GNN models without collecting sensitive data into a single computation unit. Still, privacy issues may arise as the analysis of local model updates produced by the federated clients can return information related to sensitive local data. For this reason, researchers proposed solutions that combine federated learning with Differential Privacy strategies and community-driven approaches, which involve combining data from neighbor clients to make the individual local updates less dependent on local sensitive data.

Deep neural networks (DNNs) achieve top performance through costly training on large datasets. Such resources may not be available in some scenarios, like IoT or healthcare. Extreme learning machines (ELMs) aim to alleviate this problem using single-layered networks, requiring fewer training resources. Current investigations have found that DNNs are prone to security and privacy threats, where malfunction of the network or training data extraction can be performed.

Vision Transformer (ViT) shows superior performance on various tasks, but, similar to other deep learning techniques, it is vulnerable to adversarial attacks. Due to the differences between ViT and traditional CNNs, previous works designed new adversarial training methods as defenses according to the design of ViT, such as blocking attention to individual patches or dropping embeddings with low attention. However, these methods usually focus on fine-tuning stage or the training of the model itself. Improving robustness at the pre-training stage, especially with lower overhead, has yet to be thoroughly investigated. This paper proposes a novel method, Adv-MAE, which increases adversarial robustness by masked adversarial pre-training without a penalty to performance on clean data. We design a simple method to generate adversarial perturbation for the autoencoder, as the autoencoder does not provide classification results. Then, we use masked inputs with perturbation to conduct adversarial training for the autoencoder. The pre-trained autoencoder can be used to build a ViT with better robustness. Our experimental results show that, when using adversarial fine-tuning, Adv-MAE offers better accuracy under adversarial attack than the non-adversarial pre-training method (3.46% higher on CIFAR-10, 1.12% higher on Tiny ImageNet). It also shows better accuracy on clean data (4.94% higher on CIFAR-10, 1.74% higher on Tiny ImageNet), meaning Adv-MAE does not deteriorate performance on clean inputs. In addition, masked pre-training also shows much lower time consumption at each training epoch.

Recent research has indicated that Graph Neural Networks (GNNs) are vulnerable to backdoor attacks, and existing studies focus on the One-to-One attack where there is a single target triggered by a single backdoor. In this work, we explore two advanced backdoor attacks, i.e., the multi-target and multi-trigger backdoor attacks, on GNNs: 1) One-to-N attack, where there are multiple backdoor targets triggered by controlling different values of the trigger; 2) N-to-One attack, where the attack is only triggered when all the N triggers are present. The initial experimental results illustrate that both attacks can achieve a high attack success rate (up to 99.72%) on GNNs for the node classification task.

SHA-3 is considered to be one of the most secure standardized hash functions. It relies on the Keccak-f[1 600] permutation, which operates on an internal state of 1 600 bits, mostly represented as a 5 x 5 x 64-bit matrix. While existing implementations process the state sequentially in chunks of typically 32 or 64 bits, the Keccak-f[1 600] permutation can benefit a lot from speedup through parallelization. This paper is the first to explore the full potential of parallelization of Keccak-f[1 600] in RISC-V based processors through custom vector extensions on 32-bit and 64-bit architectures. We analyze the Keccak$\mathbf{f}[1 \ 600]$permutation, composed of five different step mappings, and propose ten custom vector instructions to speed up the computation. We realize these extensions in a SIMD processor described in System Verilog. We compare the performance of our designs to existing architectures based on vectorized application-specific instruction set processors (ASIP). We show that our designs outperform all related work in throughput due to our carefully selected custom vector instructions.

Graph Neural Networks (GNNs) have achieved promising performance in various real-world applications. Building a powerful GNN model is not a trivial task, as it requires a large amount of training data, powerful computing resources, and human expertise. Moreover, with the development of adversarial attacks, e.g., model stealing attacks, GNNs raise challenges to model authentication. To avoid copyright infringement on GNNs, verifying the ownership of the GNN models is necessary.This paper presents a watermarking framework for GNNs for both graph and node classification tasks. We 1) design two strategies to generate watermarked data for the graph classification task and one for the node classification task, 2) embed the watermark into the host model through training to obtain the watermarked GNN model, and 3) verify the ownership of the suspicious model in a black-box setting. The experiments show that our framework can verify the ownership of GNN models with a very high probability (up to 99%) for both tasks. We also explore our watermarking mechanism against an adaptive attacker with access to partial knowledge of the watermarked data. Finally, we experimentally show that our watermarking approach is robust against a state-of-the-art model extraction technique and four state-of-the-art defenses against backdoor attacks.

Federated Learning (FL) has become very popular since it enables clients to train a joint model collaboratively without sharing their private data. However, FL has been shown to be susceptible to backdoor and inference attacks. While in the former, the adversary injects manipulated updates into the aggregation process; the latter leverages clients' local models to deduce their private data. Contemporary solutions to address the security concerns of FL are either impractical for real-world deployment due to high-performance overheads or are tailored towards addressing specific threats, for instance, privacy-preserving aggregation or backdoor defenses. Given these limitations, our research delves into the advantages of harnessing the FPGA-based computing paradigm to overcome performance bottlenecks of software-only solutions while mitigating backdoor and inference attacks. We utilize FPGA-based enclaves to address inference attacks during the aggregation process of FL. We adopt an advanced backdoor-aware aggregation algorithm on the FPGA to counter backdoor attacks. We implemented and evaluated our method on Xilinx VMK-180, yielding a significant speed-up of around 300 times on the IoT-Traffic dataset and more than 506 times on the CIFAR-10 dataset.

This work explores stylistic triggers for backdoor attacks in the audio domain: dynamic transformations of malicious samples through guitar effects. We first formalize stylistic triggers – currently missing in the literature. Second, we explore how to develop stylistic triggers in the audio domain by proposing JingleBack. Our experiments confirm the effectiveness of the attack, achieving a 96% attack success rate. Our code is available in https://github.com/skoffas/going-in-style.

Backdoor attacks have been demonstrated as a security threat for machine learning models. Traditional backdoor attacks intend to inject backdoor functionality into the model such that the backdoored model will perform abnormally on inputs with predefined backdoor triggers and still retain state-of-the-art performance on the clean inputs. While there are already some works on backdoor attacks on Graph Neural Networks (GNNs), the backdoor trigger in the graph domain is mostly injected into random positions of the sample. There is no work analyzing and explaining the backdoor attack performance when injecting triggers into the most important or least important area in the sample, which we refer to as trigger-injecting strategies MIAS and LIAS, respectively. Our results show that, generally, LIAS performs better, and the differences between the LIAS and MIAS performance can be significant. Furthermore, we explain these two strategies’ similar (better) attack performance through explanation techniques, which results in a further understanding of backdoor attacks in GNNs.

Abstract We present a construction of partial spread bent functions using subspaces generated by linear recurring sequences (LRS). We first show that the kernels of the linear mappings defined by two LRS have a trivial intersection if and only if their feedback polynomials are relatively prime. Then, we characterize the appropriate parameters for a family of pairwise coprime polynomials to generate a partial spread required for the support of a bent function, showing that such families exist if and only if the degrees of the underlying polynomials are either 1 or 2. We then count the resulting sets of polynomials and prove that, for degree 1, our LRS construction coincides with the Desarguesian partial spread. Finally, we perform a computer search of all $$\mathcal{PS}\mathcal{}^-$$ PS - and $$\mathcal{PS}\mathcal{}^+$$ PS + bent functions of $$n=8$$ n = 8 variables generated by our construction and compute their 2-ranks. The results show that many of these functions defined by polynomials of degree $$d=2$$ d = 2 are not EA-equivalent to any Maiorana–McFarland or Desarguesian partial spread function.

Faults occur naturally and are responsible for reliability concerns. Faults are also an interesting tool for attackers to extract sensitive information from secure chips. In particular, non-invasive fault attacks have received a fair amount of attention. One easy way to perturb a chip without altering it is the so-called Electromagnetic Fault Injection (EMFI). Such attack has been studied in great depth, and nowadays, it is part and parcel of the state-of-the-art. Indeed, new capabilities have emerged where EM experimental benches are used to cryptanalyze chips. The progress of this “field” is fast, in terms ofreproducibility,accuracy, andnumber of use-cases. However, there is too little awareness about such advances. In this paper, we aim to expose the true harmfulness of EMFI (including reproducibility) to enable reasonable security quotations. We also analyze protections (at hardware/firmware/system levels) in light of their efficiency. We characterize the specificity of EM fault injection compared to other injection means (laser, glitch, probing).

The efficiency of the profiling side-channel analysis can be significantly improved with machine learning techniques. Although powerful, a fundamental machine learning limitation of being data-hungry received little attention in the side-channel community. In practice, the maximum number of leakage traces that evaluators/attackers can obtain is constrained by the scheme requirements or the limited accessibility of the target. Even worse, various countermeasures in modern devices increase the conditions on the profiling size to break the target. This work demonstrates a practical approach to dealing with the lack of profiling traces. Instead of learning from a one-hot encoded label, transferring the labels to their distribution can significantly speed up the convergence of guessing entropy. By studying the relationship between all possible key candidates, we propose a new metric, denoted Label Correlation (LC), to evaluate the generalization ability of the profiling model. We validate LC with two common use cases: early stopping and network architecture search, and the results indicate its superior performance.

Graph Neural Networks (GNNs) are a class of deep learning-based methods for processing graph domain information. GNNs have recently become a widely used graph analysis method due to their superior ability to learn representations for complex graph data. Due to privacy concerns and regulation restrictions, centralized GNNs can be difficult to apply to data-sensitive scenarios. Federated learning (FL) is an emerging technology developed for privacy-preserving settings when several parties need to train a shared global model collaboratively. Although several research works have applied FL to train GNNs (Federated GNNs), there is no research on their robustness to backdoor attacks.

Neural networks provide state-of-the-art results in many domains. Yet, they often require high energy and time-consuming training processes. Therefore, the research community is exploring alternative, energy-efficient approaches likespiking neural networks (SNNs). SNNs mimic brain neurons by encoding data into sparse spikes, resulting in energy-efficient computing. To exploit the properties of the SNNs, they can be trained with neuromorphic datasets that capture the differences in motion. SNNs, just like any neural network model, can be susceptible to security threats that make the model perform anomalously. One of the most crucial threats is the backdoor attacks that modify the training set to inject a trigger in some samples. After training, the neural network will perform correctly on the main task. However, under the presence of the trigger (backdoor) on an input sample, the attacker can control its behavior. The existing works on backdoor attacks consider standard datasets and not neuromorphic ones. In this paper, to the best of our knowledge, we present the first backdoor attacks on neuromorphic datasets. Due to the structure of neuromorphic datasets, we utilize two different triggers, i.e., static andmoving triggers. We then evaluate the performance of our backdoor using spiking neural networks, achieving top accuracy on both main and backdoor tasks, up to 99%.

Graph Neural Networks (GNNs) have achieved impressive results in various graph learning tasks. They have found their way into many applications, such as fraud detection, molecular property prediction, or knowledge graph reasoning. However, GNNs have been recently demonstrated to be vulnerable to backdoor attacks. In this work, we explore a new kind of backdoor attack, i.e., a clean-label backdoor attack, on GNNs. Unlike prior backdoor attacks on GNNs in which the adversary can introduce arbitrary, often clearly mislabeled, inputs to the training set, in a clean-label backdoor attack, the resulting poisoned inputs appear to be consistent with their label and thus are less likely to be filtered as outliers. The initial experimental results illustrate that the adversary can achieve a high attack success rate (up to 98.47%) with a clean-label backdoor attack on GNNs for the graph classification task. We hope our work will raise awareness of this attack and inspire novel defenses against it.

Finding Boolean functions suitable for cryptographic primitives is a complex combinatorial optimization problem, since they must satisfy several properties to resist cryptanalytic attacks, and the space is very large, which grows super exponentially with the number of input variables. Recent research has focused on the study of Boolean functions that satisfy properties on restricted sets of inputs due to their importance in the development of the FLIP stream cipher. In this paper, we consider one such property, perfect balancedness, and investigate the use of Genetic Programming (GP) and Genetic Algorithms (GA) to construct Boolean functions that satisfy this property along with a good nonlinearity profile. We formulate the related optimization problem and define two encodings for the candidate solutions, namely the truth table and the weightwise balanced representations. Somewhat surprisingly, the results show that GA with the weightwise balanced representation outperforms GP with the classical truth table phenotype in finding highly nonlinear Weightwise Perfectly Balanced (WPB) functions. This is in stark contrast to previous findings on the evolution of balanced Boolean functions, where GP always performs best.

This paper uses RISC-V vector extensions to speed up lattice-based operations in architectures based on HW/SW co-design. We analyze the structure of the number-theoretic transform (NTT), inverse NTT (INTT), and coefficient-wise multiplication (CWM) in CRYSTALS-Kyber, a lattice-based key encapsulation mechanism. We propose 12 vector extensions for CRYSTALS-Kyber multiplication and four for finite field operations in combination with two optimizations of the HW/SW interface. This results in a speed-up of 141.7, 168.7, and 245.5 times for NTT, INTT, and CWM, respectively, compared with the baseline implementation, and a speed-up of over four times compared with the state-of-the-art HW/SW co-design using RV32IMC.

We consider finding as many faults as possible on the target device in the laser fault injection security evaluation. Since the search space is large, we require efficient search methods. Recently, an evolutionary approach using a memetic algorithm was proposed and shown to find more interesting parameter combinations than random search, which is commonly used. Unfortunately, once a variation on the bench or target is introduced, the process must be repeated to find suitable parameter combinations anew.To negate the effect of variation, we propose a novel method combining a memetic algorithm with a machine learning approach called a decision tree. Our approach improves the memetic algorithm by using prior knowledge of the target introduced in the initial phase of the memetic algorithm. In our experiments, the decision tree rules enhance the performance of the memetic algorithm by finding more interesting faults in different samples of the same target. Our approach shows more than two orders of magnitude better performance than random search and up to 60% better performance than previous state-of-the-art results with a memetic algorithm. Another advantage of our approach is human-readable rules, allowing the first insights into the explainability of target characterization for laser fault injection.

Finding balanced, highly nonlinear Boolean functions is a difficult problem where it is not known what nonlinearity values are possible to be reached in general. At the same time, evolutionary computation is successfully used to evolve specific Boolean function instances, but the approach cannot easily scale for larger Boolean function sizes. Indeed, while evolving smaller Boolean functions is almost trivial, larger sizes become increasingly difficult, and evolutionary algorithms perform suboptimally.

Nowadays, online services, like e-commerce or streaming services, provide a personalized user experience through recommender systems. Recommender systems are built upon a vast amount of data about users/items acquired by the services. Such knowledge represents an invaluable resource. However, commonly, part of this knowledge is public and can be easily accessed via the Internet. Unfortunately, that same knowledge can be leveraged by competitors or malicious users. The literature offers a large number of works concerning attacks on recommender systems, but most of them assume that the attacker can easily access the full rating matrix. In practice, this is never the case. The only way to access the rating matrix is by gathering the ratings (e.g., reviews) by crawling the service’s website. Crawling a website has a cost in terms of time and resources. What is more, the targeted website can employ defensive measures to detect automatic scraping. In this paper, we assess the impact of a series of attacks on recommender systems. Our analysis aims to set up the most realistic scenarios considering both the possibilities and the potential attacker’s limitations. In particular, we assess the impact of different crawling approaches when attacking a recommendation service. From the collected information, we mount various profile injection attacks. We measure the value of the collected knowledge through the identification of the most similar user/item. Our empirical results show that while crawling can indeed bring knowledge to the attacker (up to 65% of neighborhood reconstruction on a mid-size dataset and up to 90% on a small-size dataset), this will not be enough to mount a successful shilling attack in practice.

Contains fulltext : 288628.pdf (Publisher’s version ) (Open Access)

Fault injection attacks require the adversary to select suitable parameters for the attack. In this work, we consider laser fault injection and parameters like the location of the laser shot $(x,\ y)$, delay, pulse width, and intensity of the laser. The parameter selection process can be translated into an optimization problem. A very popular and successful method for various optimization problems is the genetic algorithm. To further improve the performance of a genetic algorithm, it is possible to combine it with local search to obtain a memetic algorithm. We conduct several experiments comparing the performance of the memetic algorithm and the random search algorithm for finding faults. We investigate the influence of different initialization techniques on the performance of the memetic algorithm. In our experiments, the memetic algorithm is significantly better at finding faults than the random search. While evaluating different initialization techniques, we did not observe significant differences when averaging results. However, when considering the stability of the results with a memetic algorithm based on different initialization techniques, we can distinguish preferable techniques, such as LHSMDU and the Taguchi method.

In the last few decades, evolutionary algorithms were successfully applied numerous times for creating Boolean functions with good cryptographic properties. Still, the applicability of such approaches was always limited as the cryptographic community knows how to construct suitable Boolean functions with deterministic algebraic constructions. Thus, evolutionary results so far helped to increase the confidence that evolutionary techniques have a role in cryptography, but at the same time, the results themselves were seldom used.

We investigate the use of Genetic Programming (GP) as a convolutional predictor for missing pixels in images. The training phase is performed by sweeping a sliding window over an image, where the pixels on the border represent the inputs of a GP tree. The output of the tree is taken as the predicted value for the central pixel. We consider two topologies for the sliding window, namely the Moore and the Von Neumann neighborhood. The best GP tree scoring the lowest prediction error over the training set is then used to predict the pixels in the test set. We experimentally assess our approach through two experiments. In the first one, we train a GP tree over a subset of 1000 complete images from the MNIST dataset. The results show that GP can learn the distribution of the pixels with respect to a simple baseline predictor, with no significant differences observed between the two neighborhoods. In the second experiment, we train a GP convolutional predictor on two degraded images, removing around 20% of their pixels. In this case, we observe that the Moore neighborhood works better, although the Von Neumann neighborhood allows for a larger training set.

Recently, deep learning has emerged as a powerful technique for side-channel attacks, capable of even breaking common countermeasures. Still, trained models are generally large, and thus, performing evaluation becomes resource-intensive. The resource requirements increase in realistic settings where traces can be noisy, and countermeasures are active. In this work, we exploit mimicking to compress the learned models. We demonstrate up to 300 times compression of a state-of-the-art CNN. The mimic shallow network can also achieve much better accuracy as compared to when trained on original data and even reach the performance of a deeper network.

Substitution boxes (S-boxes) are nonlinear mappings that represent one of the core parts of many cryptographic algorithms (ciphers). If S-box does not possess good properties, a cipher would be susceptible to attacks. To design suitable S-boxes, we can use heuristics as it allows significant freedom in the selection of required cryptographic properties. Unfortunately, with heuristics, one is seldom sure how good a trade-off between cryptographic properties is reached or if optimizing for one property optimizes implicitly for another property. In this paper, we consider what is to the best of our knowledge, the most detailed analysis of trade-offs among S-box cryptographic properties. More precisely, we ask questions if one property is optimized, what is the worst possible value for some other property, and what happens if all properties are optimized. Our results show that while it is possible to reach a large variety of possible solutions, optimizing for a certain property would commonly result in good values for other properties. In turn, this suggests that a single-objective approach should be a method of choice unless some precise values for multiple properties are needed.

Tasks related to Natural Language Processing (NLP) have recently been the focus of a large research endeavor by the machine learning community. The increased interest in this area is mainly due to the success of deep learning methods. Genetic Programming (GP), however, was not under the spotlight with respect to NLP tasks. Here, we propose a first proof-of-concept that combines GP with the well established NLP tool word2vec for the next word prediction task. The main idea is that, once words have been moved into a vector space, traditional GP operators can successfully work on vectors, thus producing meaningful words as the output. To assess the suitability of this approach, we perform an experimental evaluation on a set of existing newspaper headlines. Individuals resulting from this (pre-)training phase can be employed as the initial population in other NLP tasks, like sentence generation, which will be the focus of future investigations, possibly employing adversarial co-evolutionary approaches.

Side-channel attacks represent a powerful category of attacks against cryptographic devices. Still, side-channel analysis for lightweight ciphers is much less investigated than for instance for AES. Although intuition may lead to the conclusion that lightweight ciphers are weaker in terms of side-channel resistance, that remains to be confirmed and quantified. In this paper, we consider various side-channel analysis metrics which should provide an insight on the resistance of lightweight ciphers against side-channel attacks. In particular, for the non-profiled scenario we use the theoretical confusion coefficient and empirical optimal distinguisher. Our study considers side-channel attacks on the first, the last, or both rounds simultaneously. Furthermore, we conduct a profiled side-channel analysis using various machine learning attacks to recover 4-bit and 8-bit intermediate states of the cipher. Our results show that the difference between AES and lightweight ciphers is smaller than one would expect, and even find scenarios in which lightweight ciphers may be more resistant. Interestingly, we observe that the studied 4-bit S-boxes have a different side-channel resilience, while the difference in the 8-bit ones is only theoretically present.

The interplay between machine learning and security is becoming more prominent. New applications using machine learning also bring new security risks. Here, we show it is possible to reverse-engineer the inputs to a neural network with only a single-shot side-channel measurement assuming the attacker knows the neural network architecture being used.

Machine learning algorithms fall prey to adversarial examples. As profiling side-channel attacks are seeing rapid adoption of machine learning-based approaches that can even defeat commonly used side-channel countermeasures, we investigate the potential of adversarial example as a defense mechanism. We show that adversarial examples have the potential to serve as a countermeasure against machine learning-based side-channel attacks. Further, we exploit the transferability property to show that a common adversarial example can act as a countermeasure against a range of machine learning-based side-channel classifiers.

Substitution Boxes (S-boxes) are nonlinear objects often used in the design of cryptographic algorithms. The design of high quality S-boxes is an interesting problem that attracts a lot of attention. Many attempts have been made in recent years to use heuristics to design S-boxes, but the results were often far from the previously known best obtained ones. Unfortunately, most of the effort went into exploring different algorithms and fitness functions while little attention has been given to the understanding why this problem is so difficult for heuristics. In this paper, we conduct a fitness landscape analysis to better understand why this problem can be difficult. Among other, we find that almost each initial starting point has its own local optimum, even though the networks are highly interconnected.

Profiled side-channel attacks consist of several steps one needs to take. An important, but sometimes ignored, step is a selection of the points of interest (features) within side-channel measurement traces. A large majority of the related works start the analyses with an assumption that the features are preselected. Contrary to this assumption, here, we concentrate on the feature selection step. We investigate how advanced feature selection techniques stemming from the machine learning domain can be used to improve the attack efficiency. To this end, we provide a systematic evaluation of the methods of interest. The experiments are performed on several real-world data sets containing software and hardware implementations of AES, including the random delay countermeasure. Our results show that wrapper and hybrid feature selection methods perform extremely well over a wide range of test scenarios and a number of features selected. We emphasize L1 regularization (wrapper approach) and linear support vector machine (SVM) with recursive feature elimination used after chi-square filter (Hybrid approach) that performs well in both accuracy and guessing entropy. Finally, we show that the use of appropriate feature selection techniques is more important for an attack on the high-noise data sets, including those with countermeasures, than on the low-noise ones.

Finding cryptographic primitives satisfying certain properties is a difficult problem. In this domain, besides the algebraic constructions, researchers often use heuristics. There exists a set of interesting problems related to the notion of differential uniformity for a function F:\mathbbF2n→ \mathbbF2m. When n=m, then the best obtainable differential uniformity equals 2, since it is necessarily positive and even, and since examples of differentially 2-uniform functions are known. Heuristics are able to reach such functions; there is then some intuition that heuristics can be used for other open problems related to differential uniformity. When , differential uniformity is bounded by 2n-m+2 from below (when m=n-2, by 6). Unfortunately, we know such functions only for dimensions equal to n=4,5. In this paper, we explore several evolutionary algorithms and problem sizes in order to find functions having differential uniformity equal to 6. Our results show that several solution encodings are able to find such functions but only in dimensions (4, 2) and (5, 3). Since differentially 6-uniform functions were known for those sizes before, our results can be used as a source of new functions in those dimensions and as an indicator that for (6, 4) such functions either do not exist or that it is extremely difficult to find them.

Boolean functions have a prominent role in many real-world applications, which makes them a very active research domain. Throughout the years, various heuristic techniques proved to be an attractive choice for the construction of Boolean functions with different properties. One of the most important properties is nonlinearity, and in particular maximally nonlinear Boolean functions are also called bent functions. In this paper, instead of considering Boolean functions, we experiment with quaternary functions. The corresponding problem is much more difficult and presents an interesting benchmark as well as realworld applications. The results we obtain show that evolutionary metaheuristics, especially genetic programming, succeed in finding quaternary functions with the desired properties. The obtained results in the quaternary domain can also be translated into the binary domain, in which case this approach compares favorably with the state-of-the-art in Boolean optimization. Our techniques are able to find quaternary bent functions for up to 8 inputs, which corresponds to obtaining Boolean bent functions of 16 inputs.

Masking schemes represent a well-researched and successful option to follow when considering side-channel countermeasures. Still, such measures increase the implementation cost in terms of power consumption, clock cycles, and random numbers generation. In fact, the higher the order of protection against side-channel adversaries, the higher the implementation cost of countermeasures. S-boxes represent the most vulnerable part in an implementation when considering side-channel adversary. In this paper, we investigate how to generate S-boxes that have improved resilience against varying orders of side-channel attacks while having minimal implementation costs. We examine whether S-boxes generated against a certain order of attack also represent a good solution when considering different order of attacks. We demonstrate that we successfully generated S-boxes resilient against a certain physical attack order but the improvements are small. As a result, S-boxes that are resilient against first order attacks stay resilient against higher-order attacks, which saves computational power during the design of higher-order side-channel attacks resilient S-boxes.

Electromagnetic fault injection (EMFI) is a powerful active attack, requiring minimal modifications of the device under attack while having excellent penetration capabilities. The number of possible parameter combinations when characterizing an attack is usually huge, rendering exhaustive search impossible. In this work we present a novel evolutionary algorithm for optimizing the parameters for EM fault injection, which out-performs previous search methods for EMFI. The cryptographic device under attack is treated as a black box, with only a few very general assumptions on its inner workings. We test our evolutionary algorithm by attacking SHA-3 where we are able to obtain 40 times more faulty measurements and 20 times more distinct fault measurements than the random search. When coupled with the algebraic fault attack, we get 25% more exploitable faults per individual measurement.

Boolean functions and their generalizations, vectorial Boolean functions, are extremely active areas of research. Their applications can be found in domains such as error correcting codes, communication, and cryptography. Accordingly, various methods of obtaining Boolean functions are explored where one group belongs to heuristic techniques and, more precisely, evolutionary algorithms. In this paper we explore how to evolve (vectorial) Boolean functions with specific properties by utilizing several different algorithms and encodings. As far as we are aware, we are the first to explore the topic of evolution of vectorial Boolean functions where the output dimension is strictly smaller than the input dimension. Our results show that evolutionary algorithms can represent a valuable option to produce vectorial Boolean functions where good results are obtained for various sizes. On the other hand, as the number of outputs grows, we can observe that evolutionary algorithms are still able to obtain high quality results but with increasing difficulty.

We investigate the design of Orthogonal Latin Squares (OLS) by means of Genetic Algorithms (GA) and Genetic Programming (GP). Since we focus on Latin squares generated by Cellular Automata (CA), the problem can be reduced to the search of pairs of Boolean functions that give rise to OLS when used as CA local rules. As it is already known how to design CA-based OLS with linear Boolean functions, we adopt the evolutionary approach to address the nonlinear case, experimenting with different encodings for the candidate solutions. In particular, for GA we consider single bitstring, double bitstring and quaternary string encodings, while for GP we adopt a double tree representation. We test the two metaheuristics on the spaces of local rules pairs with n = 7 and n = 8 variables, using two fitness functions. The results show that GP is always able to generate OLS, even if the optimal solutions found with the first fitness function are mostly linear. On the other hand, GA achieves a remarkably lower success rate than GP in evolving OLS, but the corresponding Boolean functions are always nonlinear.

The field of side-channel analysis has made significant progress over time. Side-channel analysis is now used in practice in design companies as well as in test laboratories, and the security of products against side-channel attacks has significantly improved. However, there are still some remaining issues to be solved for side-channel analysis to become more effective. Side-channel analysis consists of two steps, commonly referred to as identification and exploitation. The identification consists of understanding the leakage and building suitable models. The exploitation consists of using the identified leakage models to extract the secret key. In scenarios where the model is poorly known, it can be approximated in a profiling phase. There, machine learning techniques are gaining value. In this paper, we conduct extensive analysis of several machine learning techniques, showing the importance of proper parameter tuning and training. In contrast to what is perceived as common knowledge in unrestricted scenarios, we show that some machine learning techniques can significantly outperform template attacks when properly used. We therefore stress that the traditional worst case security assessment of cryptographic implementations, that mainly includes template attacks, might not be accurate enough. Besides that, we present a new measure called the Data Confusion Factor that can be used to assess how well machine learning techniques will perform on a certain dataset.

The problem of obtaining maximal nonlinearity in Boolean functions is well researched, both from the cryptographic and the evolutionary computation side. However, the results are still not conclusive enough to be able to show how good a heuristic approach is when tackling this problem. In this paper, we investigate how to obtain the maximal possible nonlinearity in balanced Boolean functions, but we also analyze how difficult is the problem itself. In order to do so, we conduct experiments with Estimation of distribution algorithms as well as the fitness landscape analysis and the deception analysis. Our results indicate that the first difficulties arise from the inappropriate fitness function and representation of solutions coupled with a huge search space. The fitness landscape analysis does not reveal any significant differences that could justify the assumed jump in problem difficulty when going from Boolean functions with 6 inputs to those with 8 inputs. Finally, we show that this problem is not order-1 deceptive.

The evolution of Boolean functions that can be used in cryptography is a topic well studied in the last decades. Previous research, however, has focused on evolving Boolean functions directly, and not on general methods that are capable of generating the desired functions. The former approach has the advantage of being able to produce a large number of functions in a relatively short time, but it directly depends on the size of the search space. In this paper, we present a method to evolve algebraic constructions for generation of bent Boolean functions. To strengthen our approach, we define three types of constructions and give experimental results for them. Our results show that this approach is able to produce a large number of constructions, which could in turn enable the construction of many more Boolean functions with a larger number of variables.

The role of Boolean functions is prominent in several areas including cryptography, sequences, and coding theory. Therefore, various methods for the construction of Boolean functions with desired properties are of direct interest. New motivations on the role of Boolean functions in cryptography with attendant new properties have emerged over the years. There are still many combinations of design criteria left unexplored and in this matter evolutionary computation can play a distinct role. This article concentrates on two scenarios for the use of Boolean functions in cryptography. The first uses Boolean functions as the source of the nonlinearity in filter and combiner generators. Although relatively well explored using evolutionary algorithms, it still presents an interesting goal in terms of the practical sizes of Boolean functions. The second scenario appeared rather recently where the objective is to find Boolean functions that have various orders of the correlation immunity and minimal Hamming weight. In both these scenarios we see that evolutionary algorithms are able to find high-quality solutions where genetic programming performs the best.

Substitution Boxes (S-Boxes) play an important role in many modern-day cryptographic algorithms, more commonly known as ciphers. Without carefully chosen S-Boxes, such ciphers would be easier to break. Therefore, it is not surprising that the design of suitable S-Boxes attracts a lot of attention in the cryptography community. The evolutionary computation (EC) community also had several attempts using evolutionary paradigms to evolve S-Boxes with good cryptographic properties. This article focuses on a fitness function one should use when evolving highly nonlinear S-Boxes. After an extensive experimental analysis of the current state-of-the-art fitness functions, we present a new one that offers higher speed and better results when compared with the aforementioned fitness functions.

In data mining, decision tree algorithms are very popular methodologies since the algorithms have a simple inference mechanism and provide a comprehensible way to represent the model in the form of a decision tree. Over the past years, fuzzy decision tree algorithms have been proposed in order to provide a way to handle uncertainty in the data collected. Fuzzy decision tree algorithms have shown to outperform classical decision tree algorithms. This paper investigates a fuzzy decision tree algorithm applied to the classification of gene expression data. The fuzzy decision tree algorithm is compared to a classical decision tree algorithm as well as other well-known data mining algorithms commonly applied to classification tasks. Based on the five data sets analyzed, the fuzzy decision tree algorithm outperforms the classical decision tree algorithm. However, compared to other commonly used classification algorithms, both decision tree algorithms are competitive, although both do not reach the accuracy values of the best performing classifier.

Boolean functions are essential in many stream ciphers. When used in combiner generators, they need to have sufficiently high values of correlation immunity, alongside other properties. In addition, correlation immune functions with small Hamming weight reduce the cost of masking countermeasures against side-channel attacks. Various papers have examined the applicability of evolutionary algorithms for evolving cryptographic Boolean functions. However, even when authors considered correlation immunity, it was not given the highest priority. Here, we examine the effectiveness of three different EAs, namely, Genetic Algorithms, Genetic Programming (GP) and Cartesian GP for evolving correlation immune Boolean functions. Besides the properties of balancedness and correlation immunity, we consider several other relevant cryptographic properties while maintaining the optimal trade-offs among them. We show that evolving correlation immune Boolean functions is an even harder objective than maximizing nonlinearity.

We explore the problem space of maximum nonlinearity problems for balanced Boolean functions, examining the symmetry structure and fitness landscapes in the most common (bit string) representation. We present theoretical analyses of well understood aspects, together with detailed enumeration of the 4-bit problem, sampling of the 6-bit problem based on known optima, and sampling of the 8-bit problem based on its fittest known solutions.

Side-channel analysis (SCA) is an important issue for numerous embedded cryptographic devices that carry out secure transactions on a daily basis. Consequently, it is of utmost importance to deploy efficient countermeasures. In this context, we investigate the intrinsic side-channel resistance of lightweight cryptographic S-boxes. We propose improved versions of S-boxes that offer increased power analysis resistance, whilst remaining secure against linear and differential cryptanalyses. To evaluate the side-channel resistance, we work under the Confusion Coefficient model [1] and employ heuristic techniques to produce those improved S-boxes. We evaluate the proposed components in software (AVR microprocessors) and hardware (SASEBO FPGA). Our conclusions show that the model and our approach are heavily platform-dependent and that different principles hold for software and hardware implementations.

Security is becoming ubiquitous in our society. However, the vulnerability of electronic devices that implement the needed cryptographic primitives has become a major issue. This paper starts by presenting a comprehensive overview of the existing attacks to cryptography implementations. Thereafter, the state-of-the-art on some of the most critical aspects of designing cryptographic co-processors are presented. This analysis starts by considering the design of asymmetrical and symmetrical cryptographic primitives, followed by the discussion on the design and online testing of True Random Number Generation. To conclude, techniques for the detection of Hardware Trojans are also discussed.

Genetic algorithms are applied to numerous problems that demonstrate different properties. To efficiently solve these problems, during the years a significant number of variation operators have been and still are created. It is a problem by itself how to correctly choose between those operators, i.e. how to find the most suitable operator (or a set) for a given problem. In this paper we investigate the choice of the suitable crossover operator on the basis of fitness landscape. The fitness landscape can be described with a number of properties, so a thorough analysis needs to be done to find the most useful ones. To achieve that, we experiment with 24 noise-free problems and floating point encoding. The results indicate it is possible to either select a suitable operator or at least to reduce the number of adequate operators with fitness landscape properties.

Crossover is the most important operator in real-coded genetic algorithms. However, the choice of the best operator for a specific problem can be a difficult task. In this paper we compare 16 crossover operators on a set of 24 benchmark functions. A detailed statistical analysis is performed in an effort to find the best performing operators. The results show that there are significant differences in efficiency of different crossover operators, and that the efficiency may also depend on the distinctive properties of the fitness function. Additionally, the results point out that the combination of crossover operators yields the best results.

This paper investigates the influence of different crossover operators on the efficiency of the hybrid Taguchi genetic algorithm and aims to provide guidelines for algorithm's usage in continuous optimization. We examine the hybrid Taguchi genetic algorithm (HTGA) with 8 different crossover operators and apply it to 15 benchmark numerical optimization problems. The implementation uses binary representation which maps chromosomes to values in real domain with arbitrary precision. Different crossover operators are used with the HTGA and a detailed statistical analysis is performed to evaluate their performance. The results indicate that the HTGA obtains better results with crossover operators different than the ones commonly reported in literature.