Post Quantum Cryptography (PQC) readiness is increasingly constrained not by algorithm availability, but by cryptographic visibility, dependency complexity, and fragmented governance. This paper presents an anonymised case study of a large European critical service provider that initiated PQC readiness through a discovery first strategy, utilizing tool supported cryptographic inventorying to establish an evidence based baseline prior to migration planning. The discovery phase revealed systemic challenges, including distributed cryptographic ownership, uneven evidence quality across legacy and modern environments, and high dependency on third party cryptographic roadmaps. To operationalise these findings, the organisation introduced a structured exposure register that enabled prioritisation based on asset criticality, confidentiality longevity, and migration feasibility. We argue that PQC discovery should be understood as a governance capability that stabilises organisational knowledge and converts cryptographic uncertainty into measurable accountability, supporting risk based decision making and ecosystem coordination. The results contribute actionable lessons for institutions pursuing crypto-agility and resilience under post quantum harvest now, decrypt later threat models.

Distance-bounding (DB) protocols let a verifier upper-bound a prover’s physical distance by timing rapid challenge-response exchanges. Quantum communication promises simpler DB protocols with stronger security guarantees, yet existing quantum distance-bounding (QDB) proposals are analysed in ad hoc models and, to the best of our knowledge, lack a common game-based treatment of standard fraud attacks. We contribute (i) a reusable security framework for QDB that fixes system and timing assumptions, specifies a quantum-capable adversary model, formalises distance-, mafia-, and terrorist-fraud experiments, and includes a simple i.i.d. depolarizing noise model; and (ii) an application of this framework to a published QDB protocol. For this protocol, we characterise the honest per-round acceptance probability under noise and lift it to the multi-round setting, yielding explicit completeness guarantees as a function of the number of fast rounds, the acceptance threshold, and the noise parameter. For active adversaries, we bound the per-round success probability of distance-fraud attacks and analyse the best known mafia-fraud strategy, deriving corresponding multi-round soundness bounds. We also show that the protocol is inherently insecure against terrorist-fraud in our model. The framework cleanly separates protocol-independent definitions from protocol-specific analysis and can be used to evaluate the existing and future QDB protocols on a common basis.

Gradient-based adversarial attacks remain a dominant threat to deep neural networks (DNNs), as they exploit gradient information to efficiently optimize adversarial perturbations. To address this, we investigate whether reinforcement learning (RL) training can disrupt the gradient structure used by attackers by training image classifiers with policy-gradient objectives and epsilon-greedy exploration. Through systematic experiments across CIFAR-10, CIFAR-100, and ImageNet-100 with multiple architectures, we find that RL-trained classifiers significantly disrupt gradient-based adversarial optimization. To explain this, we conduct a comprehensive mechanism analysis using loss landscape visualization, static and dynamic gradient indicators, and predictive entropy. Our analysis reveals that RL acts as an implicit regularizer, producing models with highly unstable gradient directions and smaller gradient magnitudes. This combination makes each PGD step both unreliable in direction and limited in magnitude, causing gradient-based attacks to fail within practical iteration budgets. We further show that combining RL with adversarial training (RL-adv) provides a dual-layer defense operating at two complementary levels: RL degrades gradient information available to attackers (gradient-level defense), while adversarial training strengthens decision boundaries (boundary-level defense). RL-adv achieves the highest robustness across all major attack types evaluated, including gradient-based (PGD, AutoAttack), transfer-based, and query-based attacks, outperforming SL-adv by a significant margin. These findings identify RL-induced gradient disruption as a complementary robustness mechanism and motivate future research on hybrid SL-RL training schedules that combine SL's efficiency with RL's gradient-regularization properties.

Ranging and localisation have become critical for many applications and services. The Wi-Fi (IEEE 802.11) standard is a natural candidate for providing these functions across diverse environments, given its widespread deployment. The IEEE 802.11az amendment, finalised in 2023, introduces "Next Generation Positioning" mechanisms to secure and harden the existing insecure Wi-Fi Fine Timing Measurement (FTM) ranging solution. Moreover, the recent IEEE 802.11bk amendment increases the available bandwidth with the goal of approaching the centimetre-level ranging accuracy of ultra-wideband (UWB) systems. This paper examines to what extent these promises hold from a security and deployability perspective. We analyse the core mechanisms of secure Wi-Fi ranging as defined in IEEE 802.11az and IEEE 802.11bk at both the logical and physical layers, combining standards analysis with simulations and measurements on commercial and development hardware. At the logical layer, we show how common deployment choices can result in unauthenticated ranging, downgrade attacks, and simple denial-of-service attacks, making it difficult to securely realise many high-stakes use cases. At the physical layer, we study the predictability of secure ranging waveforms, the security impact of symbol repetition, and how waveform design choices affect compliance with spectral masks under realistic RF behaviour. Our results show that secure Wi-Fi ranging is highly sensitive to configuration choices and is non-trivial to implement on existing hardware. This is also evidenced by the currently limited support for secure Wi-Fi ranging in commodity devices. This paper provides practical guidelines for using secure FTM safely and recommendations to vendors and standardisation bodies to improve its robustness and deployability.

With the rapid growth of IoT, secure and efficient mesh networking has become essential. Thread has emerged as a key protocol, widely used in smart-home and commercial systems, and serving as a core transport layer in the Matter standard. This paper presents ThreadFuzzer, the first dedicated fuzzing framework for systematically testing Thread protocol implementations. By manipulating packets at the MLE layer, ThreadFuzzer enables fuzzing of both virtual OpenThread nodes and physical Thread devices. The framework incorporates multiple fuzzing strategies, including Random and Coverage-based fuzzers from CovFuzz, as well as a newly introduced TLV Inserter, designed specifically for TLV-structured MLE messages. These strategies are evaluated on the OpenThread stack using code-coverage and vulnerability-discovery metrics. The evaluation uncovered five previously unknown vulnerabilities in the OpenThread stack, several of which were successfully reproduced on commercial devices that rely on OpenThread. Moreover, ThreadFuzzer was benchmarked against an oracle AFL++ setup using the manually extended OSS-Fuzz harness from OpenThread, demonstrating strong effectiveness. These results demonstrate the practical utility of ThreadFuzzer while highlighting challenges and future directions in the wireless protocol fuzzing research space.

Applying Reinforcement Learning (RL) is the common approach to the realization of autonomous vehicles (AVs). In practice, this is necessarily a Multi-Agent Reinforcement Learning (MARL) given that there are multiple vehicles on the road, with each AV being an agent. This is a highly complex and dynamic operating environment, given that each vehicle has its own intent to navigate the road to safely reach its destination while avoiding accidents. In this context, the RL policies in AVs must remain robust against any adversaries to ensure road safety. It is a crucial element in the AVs’ operation on the road as they represent a cyber-physical system (CPS), i.e., any negative consequence in road safety implies potential accidents or property damage.In this paper, we experiment with adversarial examples (AEs) on vehicular communication in different phases of the RL process, and analyze their impacts on the resulting road safety. From our simulation results, we observe notable insights into various effects that AEs have in different phases of the RL process. Specifically, we note that AEs’ negative impact during the testing phase is considerably higher than during the training phase, and the conventional robustification technique in supervised learning has little effect in RL training in the AV context. This observation leads us to consider how to further approach effectively robustifying RL in autonomous driving.

Google's Fast Pair Service (GFPS) extends Bluetooth pairing with one-tap setup and account synchronisation. This paper presents the first comprehensive security analysis of GFPS. By examining 25 commercial accessories from 16 vendors across 17 unique Bluetooth chipsets, we uncover systemic enforcement failures of the specification's core security requirements. Moreover, we show that the security failures we have identified in the pairing protocol can be further cascaded, amplifying their impact across the device ecosystem. Although GFPS and Google's Find Hub network are often treated as distinct services within the broader Google ecosystem, we show that failures in one can produce severe consequences in the other. We demonstrate WhisperPair, a family of practical attacks that enables unauthorised pairing, silent hijacking of audio devices, and covert account binding that registers a victim's accessory to an attacker's account, thereby enabling persistent location tracking and stalking via Google Find Hub. These vulnerabilities are not isolated incidents but symptoms of systemic, ecosystem-wide gaps in implementation, validation, and certification. Our analysis exposes that the source of these flaws lies in GFPS's reliance on fallible, application-layer state checks rather than on cryptographic enforcement, allowing them to propagate across vendors to the end users. To address the root cause, we propose IntentPair, a lightweight protocol modification that cryptographically binds the user's pairing intent into the key schedule, eliminating the vulnerability by design. Our findings show how a small usability “add-on” can introduce large-scale security and privacy risks for hundreds of millions of users.

Understanding the security of permutation-based algorithms has been highlighted (by e.g., Naya-Plasencia and Daemen) as a central open question in symmetric cryptography. While convincing treatments have been proposed for keyed schemes, the landscape for permutation-based (keyless) hash functions remains bleak. We study permutation-based (keyless) hash functions using Rogaway’s Human Ignorance approach. We show that by adding feed-forward back to (the inner part of the state of) the sponge hash construction, the collision and (second) preimage security of the resulting construction Sponge-F can be reduced to simple and well-defined properties of the underlying permutation, i.e., correlation intractability w.r.t. a certain class of relations. The reductions hold in the quantum setting as well, enabling proving the first meaningful quantum bounds for permutation-based hash constructions using Cojocaru et al.’s lifting theorem. As a bonus, in the random permutation model, the non-quantum (second) preimage security bounds of Sponge-F are much better than the standard sponge with the same capacity.

Bitcoin’s security relies on its Proof-of-Work consensus, where miners solve puzzles to propose blocks. The puzzle’s difficulty is set by the difficulty adjustment mechanism (DAM), based on the network’s available mining power. Attacks that destroy some portion of mining power can exploit the DAM to lower difficulty, making such attacks profitable. In this paper, we analyze three types of mining power destruction attacks in the presence of petty-compliant mining pools: selfish mining, bribery, and mining power distraction attacks. We analyze selfish mining while accounting for the distribution of mining power among pools, a factor often overlooked in the literature. Our findings indicate that selfish mining can be more destructive when the non-adversarial mining share is well distributed among pools. We also introduce a novel bribery attack, where the adversarial pool bribes petty-compliant pools to orphan others’ blocks. For small pools, we demonstrate that the bribery attack can dominate strategies such as selfish mining or undercutting. Lastly, we present the mining distraction attack, where the adversarial pool incentivizes petty-compliant pools to abandon Bitcoin’s puzzle and mine for a simpler puzzle, thus wasting some part of their mining power. Similar to the previous attacks, this attack can lower the mining difficulty, but with the difference that it does not generate any evidence of mining power destruction, such as orphan blocks.

Quantum distance-bounding (QDB) protocols let a verifier check that a prover is both genuine and physically nearby. During a timed fast phase of quantum communication, the verifier measures round-trip times to obtain an upper bound on the prover's distance. For a uniform comparison, we isolate the fast phase and study one-round distance-fraud (DF) and mafia-fraud (MF) games. For discrete-variable QDB, we show that these games reduce to convex optimization problems and can therefore be solved exactly with semidefinite programming; each MF value comes with an explicit attack achieving it and a matching certificate that no attack does better. This contrasts with quantum position verification, where an attack is split between two separated parties, so its optimization is nonconvex and analyses rely on relaxations. In our MF game, the cooperating pair collapses to a single sequential strategy, which keeps the game convex and its exact value computable. Across the discrete-variable protocols we examine, the best one-round DF attack succeeds with the same probability (1/2) for every protocol, whereas MF clearly separates the protocols. For continuous-variable QDB, we report estimated attack success probabilities from a calibrated Gaussian attack model. The benchmark covers protocols whose fast phase itself authenticates the prover; designs that follow Brands and Chaum and instead bind the fast phase with a final authenticated message, like the earliest QDB proposal, fall outside it and are treated separately. Of the four protocols studied, two had no previously known one-round attack values, and we report the first ones; for the other two, we find MF attacks with higher success probability than previously reported. Overall, one-round MF resistance depends on whether an attacker can use information revealed early by the prover to answer a fresh challenge from the verifier.

The rise of Internet of Things (IoT) devices has significantly expanded the need for secure communication networks. While security testing for these protocols is crucial, there has been limited research in fuzzing cellular IoT protocol implementations. This paper presents the first framework for fuzz-testing the LTE-M and NB-IoT protocol implementations in IoT devices. Our approach focuses on testing the LTE-M and NB-IoT Attach procedure. It adapts an existing cellular protocol fuzzing framework for the specific requirements of LTE-M and NB-IoT. The framework’s effectiveness is validated through testing of four commercially deployed LTE-M/NB-IoT modems, leading to the discovery of 25 new vulnerabilities in two of the modems. This work underscores the importance of security testing emerging IoT technologies and demonstrates the potential of fuzzing to identify vulnerabilities in LTE-M and NB-IoT devices.

A Bitcoin miner who owns a sufficient amount of mining power can perform selfish mining to increase its relative revenue. Studies have demonstrated that the time-averaged profit of a selfish miner starts to rise once the mining difficulty level gets adjusted in favor of the attacker. Selfish mining profitability lies in the fact that orphan blocks are not incorporated into the current version of Bitcoin’s difficulty adjustment mechanism (DAM). Therefore, it is believed that considering the count of orphan blocks in the DAM can result in complete unprofitability for selfish mining. In this paper, we disprove this belief by providing a formal analysis of the selfish mining time-averaged profit. We present a precise definition of the orphan blocks that can be incorporated into calculating the next epoch’s target and then introduce two modified versions of DAM in which both main-chain blocks and orphan blocks are incorporated. We propose two versions of smart intermittent selfish mining, where the first one dominates the normal intermittent selfish mining, and the second one results in selfish mining profitability under the modified DAMs. Moreover, we present the orphan exclusion attack with the help of which the attacker can stop honest miners from reporting the orphan blocks. Using combinatorial tools, we analyze the profitability of selfish mining accompanied by the orphan exclusion attack under the modified DAMs. Our results show that even when considering orphan blocks in the DAM, selfish mining can still be profitable. However, the level of profitability under the modified DAMs is significantly lower than that observed under the current version of Bitcoin DAM, suggesting that orphan reporting can be an effective countermeasure against a payoff-maximizing selfish miner.

Intrusion Detection Systems (IDSs) serve as a crucial defense mechanism against cyberattacks targeting the In-Vehicle Network (IVN) of modern, interconnected vehicles. To develop and test new IDS approaches, researchers require realistic IVN data featuring real attacks on moving vehicles. To this end, this paper presents Controller Area Network and Automotive Ethernet Realistic Data Set (CarDS), a novel dataset targeting both the Controller Area Network (CAN) and Automotive Ethernet (AE) traffic of a modern, multi-domain and multi-protocol IVN. Existing datasets are often simulated or limited to basic IVN architectures consisting of only a single CAN bus. Additionally, there are no realistic datasets for AE, despite its growing importance in high-speed in-vehicle communication. CarDS addresses these limitations by providing a labeled, time-synchronized dataset of CAN and AE traces that includes both comprehensive benign profiles and sophisticated attacks. Our traces are captured from an electric vehicle from 2020 featuring a domain-oriented architecture comprising 10 internal CAN buses and 6 AE buses. Specifically, our dataset covers 9h 07m 09s of real IVN data and features 397,383,125 CAN and 180,604,377 AE messages distributed over different scenarios in 258 traces.

The accuracy and robustness of machine learning models against adversarial attacks are significantly influenced by factors such as training data quality, model architecture, the training process, and the deployment environment. In recent years, duplicated data in training sets, especially in language models, has attracted considerable attention. It has been shown that deduplication enhances both training performance and model accuracy in language models. While the importance of data quality in training image classifier Deep Neural Networks (DNNs) is widely recognized, the impact of duplicated images in the training set on model generalization and performance has received little attention. In this paper, we address this gap and provide a comprehensive study on the effect of duplicates in image classification. Our analysis indicates that the presence of duplicated images in the training set not only negatively affects the efficiency of model training but also may result in lower accuracy of the image classifier. This negative impact of duplication on accuracy is particularly evident when duplicated data is non-uniform across classes or when duplication, whether uniform or non-uniform, occurs in the training set of an adversarially trained model. Even when duplicated samples are selected in a uniform way, increasing the amount of duplication does not lead to a significant improvement in accuracy.

In recent years, vehicles have evolved into cyberphysical autonomous systems that rely on sensor data from various sources within the vehicle. With the emergence of Vehicle-to-Everything (V2X) technology, the scope of the collaborative functionality in vehicles is now expanding to the inter-vehicular level. To support these modern capabilities, the complexity of the Electronic Control Units (ECUs) and the In-Vehicle Network (IVN) architecture is rapidly increasing. As a result, IVNs are now swarms of devices that communicate safety-critical data. Unfortunately, current vehicular networks lack security, opening the path to numerous cyberattacks. A typical solution for verifying the integrity of multiple devices is swarm attestation. However, in a typical IVN setting, only the Original Equipment Manufacturer (OEM) has access to the legitimate configuration of the ECUs and does not want to disclose this information due to intellectual property and security concerns. Therefore, state- of-the-art swarm attestation schemes, which do not provide privacy guarantees, are unsuitable for IVNs.This paper proposes Secure Privacy Preserving Anonymous Swarm Attestation for In-Vehicle Networks (SPARK), which builds upon a novel group signature scheme to enable privacy-preserving, anonymous, and traceable swarm attestation of IVNs. We validate SPARK through a proof-of-concept implementation using a standardized hardware Trusted Platform Module (TPM 2.0) and representative hardware platforms. The results demonstrate the real-world applicability of SPARK.

An anonymous credential (AC) system with partial disclosure allows users to prove possession of a credential issued by an issuer while selectively disclosing a subset of their attributes to a verifier in a privacy-preserving manner. In keyed-verification AC (KVAC) systems, the issuer and verifier share a secret key. Existing KVAC schemes rely on computationally expensive zero-knowledge proofs during credential presentation, with the presentation size growing linearly with the number of attributes. In this work, we propose two highly efficient KVAC constructions that eliminate the need for zero-knowledge proofs during the credential presentation and achieve constant-size presentations. Our first construction adapts the approach of Fuchsbauer, Hanser and Slamanig (JoC'19), which achieved constant-size credential presentation in a publicly verifiable setting using their proposed structure-preserving signatures on equivalence classes (SPS-EQ) and set commitment schemes, to the KVAC setting. We introduce structure-preserving message authentication codes on equivalence classes (SP-MAC-EQ) and designated-verifier set commitments (DVSC), resulting in a KVAC system with constant-size credentials (2 group elements) and presentations (5 group elements). To avoid the bilinear groups and pairing operations required by SP-MAC-EQ, our second construction uses a homomorphic MAC with a simplified DVSC. While this sacrifices constant-size credentials (n+2 group elements, where n is the number of attributes), it retains constant-size presentations (2 group elements) in a pairingless setting. We formally prove the security of both constructions and provide open-source implementation results demonstrating their practicality. We extensively benchmarked our KVAC protocols and, additionally, bechmarked the efficiency of our SP-MAC-EQ scheme against the original SPS-EQ scheme, showcasing significant performance improvements.