In this study, we focus on sampling from the latent space of generative models built upon autoencoders so as the reconstructed samples are lifelike images. To do to, we introduce a novel post-training sampling algorithm rooted in the concept of probability mass functions, coupled with a quantization process. Our proposed algorithm establishes a vicinity around each latent vector from the input data and then proceeds to draw samples from these defined neighborhoods. This strategic approach ensures that the sampled latent vectors predominantly inhabit high-probability regions, which, in turn, can be effectively transformed into authentic real-world images. A noteworthy point of comparison for our sampling algorithm is the sampling technique based on Gaussian mixture models (GMM), owing to its inherent capability to represent clusters. Remarkably, we manage to improve the time complexity from the previous $\mathcal{O}(n \times d \times k \times i)$ associated with GMM sampling to a much more streamlined $\mathcal{O}(n \times d)$, thereby resulting in substantial speedup during runtime. Moreover, our experimental results, gauged through the Fréchet inception distance (FID) for image generation, underscore the superior performance of our sampling algorithm across a diverse range of models and datasets. On the MNIST benchmark dataset, our approach outperforms GMM sampling by yielding a noteworthy improvement of up to 0.89 in FID value. Furthermore, when it comes to generating images of faces and ocular images, our approach showcases substantial enhancements with FID improvements of 1.69 and 0.87 respectively, as compared to GMM sampling, as evidenced on the CelebA and MOBIUS datasets. Lastly, we substantiate our methodology’s efficacy in estimating latent space distributions in contrast to GMM sampling, particularly through the lens of the Wasserstein distance.

Federated Learning (FL) has been recently proposed for distributed model training at the edge. The principle of this approach is to aggregate models learned over distributed clients to obtain a new more general "averaged" model. The resulting model is then redistributed to clients for further training. To date, the most popular federated learning algorithm uses coordinate-wise averaging of the model parameters for aggregation (FedAvg). In this paper, we carry out a general mathematical convergence analysis to evaluate aggregation strategies in a FL framework. From this, we derive novel aggregation algorithms which are able to modify their model architecture by differentiating client contributions according to the value of their losses. Moreover, we go beyond the assumptions introduced in theory, by evaluating the performance of these strategies and by comparing them with the one of FedAvg in classification tasks in both the IID and the non-IID framework.

In [1], John Aycock and Tara Copplestone pose an open question, namely the explanation of the mysterious lookup table used in the Entombed Game's Algorithm for two dimensional maze generation. The question attracted media attention (BBC etc.) and was open until today. This paper answers this question, explains the algorithm and extends it to three dimensions.

This paper provides the first analysis on the feasibility of Return-Oriented programming (ROP) on RISC-V, a new instruction set architecture targeting embedded systems. We show the existence of a new class of gadgets, using several Linear Code Sequences And Jumps (LCSAJ), undetected by current Galileo-based ROP gadget searching tools. We argue that this class of gadgets is rich enough on RISC-V to mount complex ROP attacks, bypassing traditional mitigation like DEP, ASLR, stack canaries, G-Free and some compiler-based backward-edge CFI, by jumping over any guard inserted by a compiler to protect indirect jump instructions. We provide examples of such gadgets, as well as a proof-of-concept ROP chain, using C code injection to leverage a privilege escalation attack on two standard Linux operating systems. Additionally, we discuss some of the required mitigations to prevent such attacks and provide a new ROP gadget finder algorithm that handles this new class of gadgets.

In several popular standards (e.g. ISO 7816, ISO 14443 or ISO 11898) and IoT applications, a node (transponder, terminal) sends commands and data to another node (transponder, card) to accomplish an applicative task (e.g. a payment or a measurement). Most standards encrypt and authenticate the data. However, as an application of Kerckhoffs' principle, system designers usually consider that commands are part of the system specifications and must hence be transmitted in clear while the data that these commands process is encrypted and signed. While this assumption holds in systems representable by relatively simple state machines, leaking command information is undesirable when the addressed nodes offer the caller a large "toolbox" of commands that the addressing node can activate in many different orders to accomplish different applicative goals. This work proposes protections allowing encrypting and protecting not only the data but also the commands associated to them. The practical implementation of this idea raises a number of difficulties. The first is that of defining a clear adversarial model, a question that we will not address in this paper. The difficulty comes from the application-specific nature of the harm that may possibly stem from leaking the command sequence as well as from the modeling of the observations that the attacker has on the target node's behavior (is a transaction accepted? is a door opened? is a packet routed etc). This paper proposes a collection of empirical protection techniques allowing the sender to hide the sequence of commands sent. We discuss the advantages and the shortcomings of each proposed method. Besides the evident use of nonces (or other internal system states) to render the encryption of identical commands different in time, we also discuss the introduction of random delays between commands (to avoid inferring the next command based on the time elapsed since the previous command), the splitting of a command followed by n data bytes into a collection of encrypted sub-commands conveying the n bytes in chunks of random sizes and the appending of a random number of useless bytes to each packet. Independent commands can be permuted in time or sent ahead of time and buffered. Another practically useful countermeasure consists in masking the number of commands by adding useless "null" command packets. In its best implementation, the flow of commands is sent in packets in which, at times, the sending node addresses several data and command chunks belonging to different successive commands in the sequence.

This paper presents a novel fault attack against Substitution Permutation Networks. The main advantage of the method is an absence of necessity to know the exact cipher's input and output values. The attack relies only on the number of faulty cipher texts originated from the same unknown plaintext. The underlying model is a multiple bit-set or bit-reset faults injected several times at the same intermediate round state. This method can be applied against any round thus any round key can be extracted. The attack was shown to be efficient by simulation against several SPN block ciphers.

Recently, some active shielding techniques have been broken (e.g. by FlyLogic). The caveat is that their geometry is easy to guess, and thus they can be bypassed with an affordable price. This paper has two contributions. First of all, it provides a definition of the objectives of shielding, which is seldom found in publicly available sources. Notably, we precise the expected functionality, but also the constraints it must meet to be both manufacturable and secure. Second, we propose an innovative solution based on random shielding. The goal of this shielding is to make the geometry of the shield difficult to recognize, thereby making the "identification" phase of the attack harder than in previous schemes. Also, a proof of the shielding existence for two layers of metal is provided, which guarantees that the generation of the layout will succeed. Finally, we provide real tests of the shield generation algorithm, that show it is computationally tractable even for large areas to protect.

Implementations of cryptographic algorithms continue to proliferate in consumer products due to the increasing demand for secure transmission of confidential information. Although the current standard cryptographic algorithms proved to withstand exhaustive attacks, their hardware and software implementations have exhibited vulnerabilities to side channel attacks, e.g., power analysis and fault injection attacks. This paper focuses on fault injection attacks that have been shown to require inexpensive equipment and a short amount of time. The paper provides a comprehensive description of these attacks on cryptographic devices and the countermeasures that have been developed against them. After a brief review of the widely used cryptographic algorithms, we classify the currently known fault injection attacks into low-cost ones (which a single attacker with a modest budget can mount) and high-cost ones (requiring highly skilled attackers with a large budget). We then list the attacks that have been developed for the important and commonly used ciphers and indicate which ones have been successfully used in practice. The known countermeasures against the previously described fault injection attacks are then presented, including intrusion detection and fault detection. We conclude the survey with a discussion on the interaction between fault injection attacks (and the corresponding countermeasures) and power analysis attacks.

This note describes laser fault experiments on an 8-bit 0.35μm microcontroller with no countermeasures. We show that reproducible single-bit faults, often considered unfeasible, can be obtained by careful beam-size and shot-instant tuning.

The variable clock (VC) side-channel countermeasure consists in clocking a chip with an internal oscillator whose parameters (frequency, duty cycle, shape, etc.) vary randomly in time. In this paper, we use parametric deconvolution to process VC-power consumption curves. We also analyze experimental results in order to show its efficiency.

A variant of Waters' identity-based encryption scheme with a much smaller system parameters size (only a few kilobytes) is presented. It is shown that this variant is semantically secure against passive adversaries in the standard model. In essence, the new variant divides Waters' system parameters size by a factor ℓ at the cost of (negligibly) reducing security by ℓ bits. The construction yields a fully secure practical identity-based encryption scheme.

The effect of faults on electronic systems has been studied since the 1970s when it was noticed that radioactive particles caused errors in chips. This led to further research on the effect of charged particles on silicon, motivated by the aerospace industry, which was becoming concerned about the effect of faults in airborne electronic systems. Since then various mechanisms for fault creation and propagation have been discovered and researched. This paper covers the various methods that can be used to induce faults in semiconductors and exploit such errors maliciously. Several examples of attacks stemming from the exploiting of faults are explained. Finally a series of countermeasures to thwart these attacks are described.

In addition to its usual complexity assumptions, cryptography silently assumes that information can be physically protected in a single location. As one can easily imagine, real-life devices are not ideal and information may leak through different physical channels.This paper gives a rigorous definition of leakage immunity and presents several leakage detection tests. In these tests, failure confirms the probable existence of secret-correlated emanations and indicates how likely the leakage is. Success does not refute the existence of emanations but indicates that significant emanations were not detected on the strength of the evidence presented , which of course, leaves the door open to reconsider the situation if further evidence comes to hand at a later date.

This paper introduces a simple alternative to the hash-and-sign paradigm, from the security point of view but for signing short messages, called twinning. A twin signature is obtained by signing twice a short message by a signature scheme. Analysis of the concept in different settings yields the following results:

This paper describm a new pub~c-key cryptosystem based on the hardnxs of computing higher residuw modulo a composite MA integer.We introduce two versions of our scheme, one deterministic and the other probabi~stic.The deterministic version is practically oriented encryption amounts to a single exponentiation w.r.t. a modulus with at least 768 bits and a 160-bit exponent.Decryption can be suitably optimized so as to become less demanding than a couple RSA decryptions.Although slower than MA, the new sdeme is still reasonably competitive and has several specific applications.The probabilistic version exhibits an homomorphic encryption scheme whose expansion rate is much better than previously proposed such systems.Furthermore, it has se mantic security, relative to the hardness of computing higher residu~s for suitable moduE.

The signature generation phase of most DLP-based signature schemes (for instance Schnorr[10], El-Gamal[4] or the newly standardized D.S.A.[3]) includes the timeconsuming computation of r = g K mod p where k is random.This paper introduces a new computational strategy that can apply in this particular context :A batch exponentiation technique which allows the generation of large sets of exponentials without introducing any bias between the ks (that is, the signer can batch-compute the exponentials corresponding to arbitrarily imposed powers -for instance by an external random number generator).Our method offers real improvements over the prior art with various time and memory trade-offs.

In the magical art of Steganography, there is nothing frivolous, nor contrary to the Gospels and the Catholic faith; nor have we taught superstitious beliefs. Everything is based on natural, lawful and honest principles; the mystery which veils the precepts of this art and the names of the spirits, requires a cultivated reader; to hide the secrets of this art, which could be harmful if made known to wicked men, we avail ourselves of the services of the spirits. Johannes Trithemius, Steganographia The air is influenced by astral emanations. So one can, naturally and without spiritual help, communicate his thoughts to another man, however large the distance between them. This I have seen done, I did myself and was done by Trithemius… In the same way, one can broadcast in the air any image, however far, by means of mirrors… The image will be, through large distances, seen by a conscious reader in the lunar disc; this artifice was used by Pythagoras. H.C Agrippa De Occulta Philosophia It's better to have loved and lost than to have liked and tied for second Anonymous