The Zen of Bluetooth Security (ZOBS) is a collection of Bluetooth security principles developed by the speaker to systematize seven years of research on Bluetooth security protocols. The talk explores Bluetooth security research through the lens of these principles, distills key lessons learned, and shares them with the broader wireless-security research community. The talk also pinpoints open research challenges in Bluetooth security and opens the floor for discussion of future research directions. The audience will learn about the state of the art in Bluetooth security and how to apply the ZOBS principles to offensive and defensive wireless-security research. The keynote covers eight Bluetooth BR/EDR, commonly called Bluetooth Classic (BC), and Bluetooth Low Energy (BLE) security papers co-authored by the speaker: BlueBrothers [8], HardaBLE [7], BLERP [6], BLUFFS [1], BLUR [5], BIAS [3], KNOB on BLE [4] and KNOB on BC [2].
In 2012, the OpenSky Network (https://opensky-network.org) began as an academic experiment: a crowdsourced sensor network for studying the security of wireless aviation communication. At the time, ADS-B and related surveillance technologies were becoming central to modern aviation, yet their openness, lack of authentication, and real-world deployment at scale raised fundamental security questions.The questions that were posed by several collaborating academic security groups were: What would it take to observe, measure, and understand these systems not in a laboratory, but in the wild? What is required to scale a large distributed sensor network?More than a decade later, OpenSky has become something much larger than its founders originally imagined. Built by researchers, aviation enthusiasts, and citizen scientists, it has grown into a global non-profit infrastructure for collecting, archiving, and sharing open aviation data. Its database now contains tens of trillions of records across petabytes of data, and has supported more than 900 scientific publications in many diverse fields including aviation security, air traffic management, climate science, pandemic modelling, machine learning, economics, journalism, and public accountability. Beyond academia, it has been used by many diverse partners from EUROCONTROL to the Bank of England.This keynote tells the story of how a security research project became global infrastructure. We reflect on the technical, scientific, organizational, and ethical lessons learned while operating a long-running, crowdsourced, non-profit data collection system. We will revisit OpenSky's original motivation in aviation security, including insecure aircraft communications and later work on GNSS interference and spoofing, but also show how the same infrastructure enabled discoveries far beyond wireless security.The talk discusses the power and fragility of crowdsourced measurement: how to build trust in noisy global data, how to sustain a volunteer-based network over many years, how to balance openness with privacy and safety, and how to keep a public-good infrastructure alive in a world increasingly dominated by proprietary data platforms. The broader message is that security research infrastructure can have second lives: when built openly, carefully, and sustainably, it can become a foundation for science, society, and unexpected forms of public value.
The Final Security Frontier: Using Privacy-Preserving Computation to Secure Satellite Rendezvous and Proximity Operations
Space is emerging as a critical domain for secure and privacy-preserving computing, driven by the rapid growth of commercial satellites and the increasing complexity of in-space operations. This need is particularly evident in satellite rendezvous and proximity operations (RPO), where multiple satellites must compute with private, identifying, or even classified information in order to operate safely at close distances. Secure multiparty computation (MPC) offers a promising foundation for privacy-preserving collaboration, but its suitability for in-space applications remains largely unexplored. This work provides a domain-informed analysis of satellite RPO and private computation techniques for two fundamental scenarios: collision avoidance and multi-point inspection. We design and evaluate a secure two-party computation for collision avoidance and a three-party computation for multi-point inspection using the MP-SDPZ compiler. Testing on radiation-tolerant NVIDIA Jetson hardware, we find that collision avoidance can be performed securely in 7.38 seconds in a semi-honest dishonest majority model, while multi-point inspection can be performed securely in 0.28 seconds in the semi-honest model and in 2.7 seconds in the malicious model. These findings demonstrate that MPC is both feasible and practical for privacy-preserving RPO under realistic hardware and threat assumptions. To our knowledge, this is the first work to provide granular assessment and experimentally validated MPC for in-space cooperative operations, opening a new direction for cybersecurity research in emerging space systems.
Denied by Border: Denial-of-Service Attack Exploiting Location Restrictions in Non-Terrestrial Networks
Non-Terrestrial Network (NTN) is an emerging technology aimed at extending cellular connectivity on a global scale via satellite constellations. Because satellite beam footprints can span national borders, 3GPP specifications define location-based restrictions to comply with regional regulations. When the network infers that a user equipment (UE) is located in a region where the UE's mobile operator is not permitted to provide service, it issues a network rejection to block access. In this paper, we show that this location-based authorization mechanism contains a specification vulnerability that enables a relay-based Denial-of-Service (DoS) attack. We present the Denied by Border attack, where a Man-in-the-Middle (MitM) attacker redirects the victim's signaling through a relay endpoint located in a forbidden area, causing the network to perceive the access attempt as originating from a restricted location. To demonstrate the feasibility of this attack, we evaluate its security impact on three commercial UEs and find that it results in prolonged DoS conditions on all tested devices, further impacting terrestrial network access. These findings highlight the security risks inherent to location-based access control in NTNs and motivate the need for robust location verification in future standards. The identified vulnerability has been acknowledged by the GSMA.
Satellite systems are increasingly vulnerable to spoofing attacks at the physical layer, where adversaries use inexpensive radio equipment to interfere with and replace legitimate signals. While cryptographic countermeasures are common in other wireless systems, their adoption in new space programs is slow due to concerns about the associated implications on robustness, cost, weight, power, and the challenges of updating existing systems. In this paper we introduce SideDish, a novel anti-spoofing countermeasure that combines a secondary receiver colocated at the satellite receiver with decoded signal comparison to detect out-of-beam unauthentic interference. The system is retrofittable into existing ground station deployments, cheap by using only low-cost components, and is robust against denial of service attacks. We verify this through simulations and real-world experiments that show SideDish spatially constraints attackers by between 70-99.84% in the angular domain, even considering scattering effects of the primary antenna. Targeting SideDish to deny service is not feasible within practical constraints, requiring microsecond-order timing accuracy to overcome.
Modern homes are transformed into highly connected environments through the integration of various Internet of Things (IoT) devices. This rapidly expanding ecosystem significantly reshapes the security landscape, introducing substantial privacy risks. Limited computational resources and the lack of consistent security standards contribute to the emergence of exploitable vulnerabilities, exposing users' privacy, particularly during the provisioning phase. In this paper, we conduct an in-depth analysis of commonly deployed IoT devices by examining cryptographic schemes, pairing mechanisms, and communication phases. We propose a systematic methodology to analyze the exploitability of provisioning vulnerabilities by an adversary, along with þetool. Available at https://gitlab.inria.fr/fun-team/pmfinspect, an automated tool that verifies PMF implementation and evaluates device robustness against forged management frames. Applied to 25 IoT devices from 10 different vendors, our evaluation shows that 4/10 vendors rely on weak protocols, 3/10 do not implement PMF, and 18/25 devices are vulnerable to combined attacks exploiting both weaknesses. We finally identify the underlying causes and discuss mitigation strategies.
The Cost of Zero Trust: A Comparative Analysis of MACsec and IPsec Architectures for Secure Open Fronthaul
The Open Fronthaul (OFH) interface requires strict latency (> 100 μs) and precise synchronization, creating a critical trade-off between security and network performance on Commercial Off- The-Shelf (COTS) hardware. In contrast to prior art relying on synthetic traffic or ignoring cross-plane interference, we systematically evaluate Media Access Control Security (MACsec), kernelspace Internet Protocol Security (IPsec), and Data Plane Development Kit (DPDK)/Vector Packet Processing (VPP) IPsec on a Functional Split 7.2x testbed using realistic, frame-synchronous 5G New Radio (NR) workloads. We address three key challenges: (i) determining whether encryption protocols can meet strict Open RAN (O-RAN) timing budgets; (ii) identifying encryption solutions that enable cloud-based, multi-hop deployments while complying with these timing constraints; and (iii) maintaining nanosecondprecision Precision Time Protocol (PTP) synchronization under heavy cryptographic loads on commodity COTS hardware. Our results demonstrate that while MACsec achieves excellent timing performance (44 μs jitter), it requires Layer 2 (L2) adjacency, precluding multi-hop cloud deployments. Conversely, kernel IPsec enables topology hiding but introduces prohibitive jitter (580-590 μs), violating O-RAN timing budgets. We demonstrate that DPDK/VPP IPsec bridges this gap, achieving a 10x jitter reduction (45-66 μs) comparable to MACsec, while enabling routed security. Furthermore, we show that PTP remains stable (∼ 27 ns jitter) under secured loads up to 5.75 Gbps using kernel-space IPsec and 10 Gbps using DPDK/VPP IPsec, provided hardware timestamping is enabled. We conclude that the IPsec performance bottleneck stems principally from kernel stack traversal, establishing DPDK/VPP-based kernel-bypass IPsec as a practical, software-defined path to secure OFH.
The rapid development of large language models (LLMs) has led to their increasing deployment on edge devices for applications such as autonomous driving. These edge deployments enable efficient localized processing while enhancing user privacy. However, unintentional leaks of model families and hyperparameters during LLM execution serve as critical attack vectors, facilitating high-level threats like model extraction and membership inference. In this paper, we propose ArchSnoop, which is a non intrusive edge LLM architecture eavesdropping attack based on electromagnetic (EM) leakage. Our key insight is that the EM signals emitted by GPU and memory activities during inference reflect hierarchical architectural features ranging from token generation to individual linear projections. We designed a two-stage hierarchical reconstruction model to recover fine-grained architectural information from these EM signals. Our evaluation on the NVIDIA Jetson Orin Nano platform demonstrates that ArchSnoop achieves high accuracy in architecture reconstruction, including 99.12% in model family classification and 97.11% for hyperparameter estimation. We reveal the mapping between EM signals and LLM architectures on edge devices for the first time. By proposing potential countermeasures such as physical EM shielding and software-level perturbations, this work provides a new dimension to secure edge computing platforms against physical threats.
Many existing Artificial Intelligence (AI) solutions on mobile devices rely on an extensive collection of sensitive data, raising privacy concerns and often requiring storage for both context and model improvement. Apple's Private Cloud Compute (PCC) aims to address this by emphasizing mobile device integration and a privacy-first design. The central claim of PCC is that it does not store any user data and that user input and user accounts are unlinkable.While most of the PCC system specifications are public, compiled binaries add a layer of opaqueness. There are no reproducible builds, and there are no symbols within those binaries, creating potential discrepancies between the specification and what is shipped to the user. Additionally, the underlying models and interfaces for querying PCC are not openly accessible, limiting academic evaluation of model properties, such as accuracy. This poses a challenge in assessing whether a privacy-preserving approach like PCC is actually trustworthy while also providing high-quality answers.We are the first to reverse-engineer the PCC implementation on mobile devices to evaluate privacy aspects and to open its non-public interfaces on local devices to support custom PCC queries. We demonstrate this level of access beyond Apple's intended use cases by independently benchmarking the PCC model. We enable future research by making our PCC benchmarking framework publicly available.
Pairing of wireless devices, such as smartphones, suffers from a plethora of practical challenges if strong security guarantees via integrity codes are desired. Acoustic schemes are easy to deploy, but suffer from limited data rates and high sensitivity to environmental noise. Existing Wi-Fi-based methods, however, lack the necessary timing precision and signal flexibility to implement integrity codes on off-the-shelf devices.In this paper, we introduce Pair-Fi, a novel method that overcomes these limitations by leveraging integrity codes transmitted using software-defined radio (SDR)-like capabilities enabled through firmware modifications directly on smartphone-integrated Wi-Fi chips. Pair-Fi demonstrates both transmission and reception of raw IQ samples directly on commodity smartphone hardware, allowing for precise on-off keying modulation with timing resolution as low as 4 µs slots. By bypassing the constraints of conventional Wi-Fi frame timings, our approach significantly improves pairing speed, reliability, and resistance to interference. We validate Pair-Fi experimentally on recent smartphone models, such as the Google Pixel 7, showing robust performance in realistic environments. Our results indicate that SDR-enabled integrity code pairing via smartphone Wi-Fi chips provides a practical, secure, and efficient alternative to existing device pairing mechanisms, opening new avenues for secure and seamless device interactions in everyday scenarios.
Bluetooth Low Energy (BLE) is a ubiquitous wireless technology used by billions of devices and defined in an open standard. The BLE specification defines two security protocols: pairing, which establishes a trust relationship between two devices by deriving the Long-Term Key (LTK), and session establishment, which generates a fresh encryption key for each (re)connection. The BLE security model and prior research primarily consider wireless-only adversaries. However, real deployments increasingly face software compromise, where an attacker exploits a vulnerability to gain arbitrary code execution or memory read/write capabilities on the device. Under such a compromise, an attacker can extract the LTK and use it to impersonate trusted devices or decrypt/forge protected traffic.To address the gap between BLE's wireless-only assumptions and practical system security, we present HardaBLE, a hardened BLE architecture designed to protect keys and prevent their use under software compromise. HardaBLE confines LTK storage and LTK-dependent cryptographic operations to a hardware-isolated secure environment, preventing key exfiltration. Furthermore, it enforces integrity-bound authorization by denying LTK-dependent operations if the secure environment cannot verify firmware integrity evidence against a stored reference. We formally model HardaBLE in TLA+ and verify it with TLC, demonstrating that our design prevents key exfiltration and denies key-dependent operations when firmware integrity evidence is invalid.We implement HardaBLE by modifying the Zephyr BLE stack, and we evaluate it on nRF53 development boards. We also empirically validate our prototype under simulated software compromise, showing that it does not expose keys in plaintext and denies key-dependent operations when firmware integrity evidence verification fails. Our performance evaluation shows that hardware-isolated key handling adds negligible latency and energy overhead. Enabling integrity verification increases session establishment time by up to 18% and approximately doubles the per-reconnection charge-draw. However, the absolute overhead remains under 0.3 microampere-hours (µAh), making HardaBLE still practical for common BLE workloads with infrequent reconnections.
In Enterprise Wi-Fi networks such as eduroam, clients typically verify the network's authenticity by validating the authentication server's certificate. To learn this certificate without relying on pre-configuration, many operating systems rely on a non-standardized TOFU mechanism. In this work, we empirically analyse how operating systems implement TOFU, and find that they are inconsistent and flawed. Notably, no tested client validates the certificate's expiry of trusted networks, and Linux lacks TOFU entirely, in practice making Linux users less secure. A root cause of these issues is the lack of a standardized TOFU design. To remedy this, we propose design guidelines for TOFU, a scheme to automatically use anonymous identities during TOFU, and a mechanism that avoids relying on TOFU altogether, where mutual trust is instead established based on the user's password. We demonstrate the practicality of our design through an implementation on top of Linux. During this, we patched certificate validation flaws in Linux's
Bluetooth is a pervasive wireless standard that, despite numerous revisions, remains vulnerable to multiple design-level security flaws. Specifically, its pairing and session establishment security protocols lack integrity protection, forward secrecy, or strong authentication mechanisms, thereby enabling critical impersonation and man-in-the-middle attacks. These risks are compounded by complex and fragmented specifications, which hinder secure implementation and formal analysis.To address these issues, we present BlueBrothers, three new protocols to serve as a secure alternative to the current ones. BB-Pairing combines pairing and session establishment in a single protocol that provides integrity protection and robust user-assisted authentication. BB-Session establishes authenticated, secure sessions with forward secrecy guarantees. BB-Rekey provides forward and future secrecy within a session via a lightweight key-refresh mechanism.We model BlueBrothers in ProVerif and verify confidentiality, integrity, and entity-authentication properties. We implement the protocols on constrained nRF52 devices and evaluate performance against the Bluetooth baseline. Our results show up to a 59% reduction in latency with comparable energy consumption.
Detecting malicious Bluetooth Low Energy (BLE) trackers remains challenging because existing approaches rely on fixed time and distance thresholds that are brittle across environments. These heuristics produce false positives in dense environments and require long observation windows before flagging a device. To address these limitations, we present BL(u)E CRAB, a cross-platform (iOS/Android) mobile system that represents nearby devices using three risk factors derived from BLE scan data. Our detection model adapts Clustering-Based Local Outlier Factor (CBLOF) to BLE tracker detection and adds a gap-thresholding mechanism to separate high-scoring outliers from the benign majority. Across micro-benchmarks and end-to-end case studies, CBLOF reduces false positives up to 77% and false negatives up to 20% compared to the state of the art. In our case studies, suspicious trackers are typically detected within 5 minutes of scanning, improving practical usability for real-world deployment.
Logging is a common practice in software development, widely used for debugging, testing, and performance monitoring. However, recording sensitive user data can introduce severe privacy risks. Past incidents involving leaked logs have prompted platforms such as Android to publish strict guidelines discouraging developers from logging personally identifiable information (PII) and other ''linkable'' or ''ambiguous'' data unless strictly required for core functionality. To evaluate real-world compliance, we examined the logging practices of 500 Android applications across six categories. Our findings reveal that 264 apps contain logging violations, from which we identified 864 instances of sensitive data exposure. Notably, 54% of these violations stem from debugging logs that should have been removed before release. The recorded data includes PII such as email addresses and phone numbers, as well as linkable information such as shopping history, health records, and private messages each in direct violation of Android's privacy guidelines. Moreover, removing these logging statements from application source code does not affect app functionality, raising questions about their necessity. Our analysis shows that most violations originate from debugging practices, third-party analytics tracking, and HTTP request logging. Further, by applying a large language model (LLM) to inspect an additional set of 300 applications, we found 240 apps exhibiting sensitive data logging violations. We also discovered that many apps share log data with third-party services, often contradicting their own privacy policies. To mitigate these risks, we provide practical recommendations for both app developers and mobile platforms to enforce responsible and privacy-preserving logging practices.
Sideloading apps in iOS is possible without a jailbroken (rooted) device despite its 'walled garden' reputation. We found 89 third-party app markets and present the first overview of the modded app ecosystem on iOS. We collected and analysed apps from the 9 most popular modded markets over 11 months. We found 29% of the modded apps are pirated paid apps offered for free with a reported 1.12 billion downloads across just three markets, likely affecting developer revenue. Many modded apps offer subscription features and in-app or in-game items for free: 69% of modded apps originally had in-app purchases in the App Store. In terms of user security, VirusTotal classified over 1% of modded apps as malicious (CVE, exploits, trojans, mainly), versus less than 0.03% of App Store apps. Markets sign 78% of the modded apps with enterprise certificates, while others rely on device management profiles meant for company devices which can extract private user data and remotely install apps.
Mobile apps offer significant benefits, but their privacy protections often remain ineffective and confusing for users. While prior work mainly analyzes app privacy vulnerabilities, few approaches help users understand, set, and enforce their privacy preferences. This paper presents PrivacyAssist, a multi-agent LLM-based platform that detects inconsistencies between user-granted permissions and developers' declared sensitive data collection and sharing practices. Using Retrieval-Augmented Generation (RAG), PrivacyAssist provides concise explanations and real-time on-device warnings to support informed installation decisions. We evaluate PrivacyAssist with 200 users and 2,347 Android apps, finding that only 16% of apps are fully consistent between granted permissions and declared data practices.
Google's Play Integrity framework has replaced SafetyNet as the primary mechanism for app, device, and account attestation on Android, yet its real-world adoption remains poorly understood. We present AttestLens, a large-scale measurement of attestation usage across 125,421 APKs collected from AndroZoo in April 2025. Our dataset spans six app categories—popular, banking, cryptocurrency, government, gaming, and random—and includes a longitudinal set of 814 government applications.We develop robust static markers to detect SafetyNet and Play Integrity, even under common obfuscation techniques, and use them to quantify adoption by category and by app characteristics such as downloads, release date, and rating. Our findings are concerning: roughly 90% of apps in every category reference neither framework, and adoption remains particularly low in security-sensitive areas such as banking, cryptocurrency, and gaming. Additionally, many government apps continue to reference SafetyNet despite its deprecation. Across 6,281 Play Integrity-invoking applications, we find that most attestation originates from third-party packages rather than first-party invocation. On a positive note, most invoking apps follow Google's recommendation of pairing Play Integrity with other defense-in-depth techniques.
The open radio access network (O-RAN) standard provides a foundational move towards disaggregated RAN architecture, allowing flexibility and multi-vendor integration. For example, the Radio Intelligent Controller (RIC) may involve third-party applications (xApps) to dynamically control and monitor network behavior, facilitating significant opportunities for multi-party involvement, but allowing potentially untrusted integration with the RAN. In this paper, we propose, design and evaluate
5G promises enhanced performance—not only in bandwidth and capacity, but also latency and security. Its ultra-reliable low-latency configuration targets round-trip times below 1 ms, while optional security controls extend protection across all interfaces, making 5G attractive for mission-critical applications. A key enabler of low latency is the disaggregation of network components, including the RAN, allowing user-plane functions to be deployed nearer to end users. However, this split introduces additional interfaces, whose protection increases latency overhead. In this paper, guided by discussions with a network operator and a 5G manufacturer, we evaluate the latency overhead of enabling optional 5G security controls across internal RAN interfaces and the 5G user plane. To this end, we deploy the first testbed implementing a disaggregated RAN with standardized optional security mechanisms. Our results show that disaggregated RAN deployments retain a latency advantage over monolithic designs, even with security enabled. However, achieving sub-1 ms round-trip times remains challenging, as cryptographic overhead alone can already exceed this target.
The ongoing digital transformation towards Sixth Generation (6G) networks is driven by connected intelligence, enabling multi-sensory experiences, brain–computer interaction, and eHealth applications. This evolution has led to an explosion of Internet of Things (IoT) devices, many of which remain vulnerable and are frequently exploited to launch Internet Protocol (IP)-based Distributed Denial of Service (DDoS) attacks, designed to overwhelm their targets' resources, causing service disruptions. While IP-DDoS attacks are widely studied, their detection is limited to scrutinizing packets at the IP layer in the core network, thus overlooking the need for their early detection at the Radio Access Network (RAN). In this work, we investigate the manifestation of IP-based DDoS attacks at the Physical (PHY) and Medium Access Control (MAC) layers under different User Equipment (UE) mobility patterns, showcasing that features collected from these layers can be leveraged to detect attacks occurring in higher layers of the protocol stack, even when their payloads are encrypted. Additionally, we propose RAN-GUARD, a novel hybrid multi-model approach that combines a Long Short-Term Memory (LSTM) AutoEncoder (AE) with an Isolation Forest (IF) to detect IP-based DDoS attacks using features from only the MAC and PHY layers. RAN-GUARD achieves an average F1-score of 95.5% across four different IP-based DDoS attacks and an average F1-score of 92% across five mobility patterns, outperforming the state-of-the-art by an average of 19.4%. Furthermore, RAN-GUARD leverages the Shapley Additive Explanations (SHAP) technique, which shows that PortScan, DDoS Ripper, and DoS HULK attacks are exacerbated through MAC-based features, while PHY layer features can assist the MAC ones in detecting Slowloris attacks.
5G networks provide low-latency, high throughput, and massive connectivity, yet the control plane remains exposed to several security threats. Among the most common and impactful threats are Denial-of-Service (DoS) attacks, with Radio Resource Control (RRC) signaling storms being particularly effective and difficult to mitigate. In this attack, a malicious User Equipment (UE) aims to exhaust Next Generation Node Base (gNB) resources, preventing legitimate UEs from establishing a connection. Existing defenses are typically limited to detection, only evaluated through numerical simulations, and cannot discern between high-load network conditions and attacks. Most of them also assume static setups and do not take mobility into account. In this paper, we first evaluate the feasibility of the signaling storm attack by using the OpenAirInterface (OAI) 5G protocol stack. Then, we propose StormShield, a signaling storm attack detection and mitigation technique implemented as an xApp on an O-RAN Near-Real-Time (near-RT) RAN Intelligent Controller (RIC). It fingerprints and blocks Malicious UEs (MUEs) before gNB resources are exhausted. We prototyped our solution on an Over-The-Air (OTA) testbed with OAI, NVIDIA Aerial, and two different gNB setups. The first one leverages an USRP X410 Software-defined Radio (SDR) with 8.1 functional split; the second a commercial Foxconn Radio Unit (RU) with 7.2 functional split. Our experimental evaluation demonstrates that StormShield effectively prevents gNB resource exhaustion, identifying and blocking MUEs with an average detection accuracy of 97.6% within 106.5 ms from the beginning of the attack.
MUTUALISM: Low-Footprint and High-Throughput Software Implementation of HQC for Resource-Constrained Devices
Hamming Quasi-Cyclic (HQC) is a code-based key-encapsulation mechanism selected by NIST for post-quantum cryptography standardization, but it is challenging to deploy on resource-constrained devices due to its memory and computational costs. In this paper, we propose a novel software implementation of HQC, named MUTUALISM, which is deployable on resource-constrained devices. MUTUALISM builds on an efficient multiply-then-reduce method for sparse-dense polynomial multiplication that minimizes costly modular reduction operations. We further propose two variants of the multiplication algorithm that optimize memory footprint and computation throughput, respectively. We then present MUTUALISM as a unified and optimized implementation derived from fine-grained profiling of HQC. This implementation systematically integrates the aforementioned multiplication algorithms with additional optimizations throughout the HQC implementation pipeline, thereby reducing the memory footprint and improving end-to-end throughput. Our experiments on ARM Cortex-M4 show that MUTUALISM achieves a speedup of more than 57 times compared with PQClean while reducing the memory footprint by 80%. To the best of our knowledge, MUTUALISM is the first software-only HQC implementation deployable on resource-constrained devices with 64 KB of RAM across all security levels. This enables post-quantum key establishment on BLE-class IoT microcontrollers with only tens of kilobytes of SRAM, thereby bridging the gap between PQC standards and real-world wireless embedded deployments.
TinyContainer: Container Runtime Middleware Enabling Multi-tenant Microcontrollers with Built-in Security
Software containerization technologies for resource-limited devices enable multi-tenant microcontrollers, which allow running multiple applications with different permission levels. However, current solutions lack run time configuration over various settings on container scheduling and container permissions to host resources. This limits the applicability of constrained containerization in dynamic and heterogeneous environments. This paper introduces TinyContainer, a lightweight software container management middleware designed for multi-tenant microcontrollers. TinyContainer provides per-container configurable scheduling and fine-grained access control to host resources through a metadata-driven approach, supporting multiple runtimes via a runtime abstraction layer. We analyze the performance of TinyContainer with a small WebAssembly runtime, CS4WAMR, and RIOT OS, a common RTOS. We report on experiments using popular IoT boards based on various Cortex-M microcontrollers. We show the endpoint system brought by TinyContainer allowing to regulate access of containers to host resources and provide host services to containers with an overhead of up to 4 ms per call. In particular, we showcase a TinyML use case, whereby containers retain data and model weights, while model inference is delegated to native host RTOS services.
Standardized Vehicle-to-everything (V2X) security architectures, such as the Secure Credential Management System (SCMS) in the U.S. and the C-ITS Security Credential Management System (CCMS) in Europe, protect driver privacy through large-scale batch issuance of short-lived pseudonym certificates. At national deployment scale, minute-level rotation across hundreds of millions of vehicles results in trillion-scale annual certificate management, creating substantial backend complexity, storage burden, and persistent cellular connectivity requirements.This paper presents V-PASS, a hardware-rooted architecture that enables vehicles to derive unlinkable pseudonyms locally from a single enrollment credential, eliminating periodic batch provisioning from the Vehicular PKI (VPKI). A Secure Element (SE) enforces a window-level authorization policy and produces a hardware-bound attestation, ensuring that at most one valid pseudonym can be derived per window even under ECU software compromise, thereby providingV-PASS adopts a split-trust design that confines lightweight authorization to the SE while delegating computationally expensive cryptographic proofs to the ECU. The resulting window object enables stateless verification for receivers while preserving cross-window unlinkability and conditional accountability.We implement V-PASS on a Raspberry Pi 5 emulating constrained automotive hardware. Evaluation shows a verification latency of 0.30 ms per message, supporting real-time V2X safety messaging, while achieving a 1750× reduction in pseudonym storage compared to traditional VPKIs, demonstrating practical feasibility for large-scale deployment.
Radio Frequency Fingerprinting (RFF) has emerged as a promising physical-layer technique for device identification, leveraging the hardware imperfections in radio transmitters. However, the assumption that RF fingerprints are static and persistent is increasingly challenged by recent findings. In this work, we present a comprehensive statistical analysis of radio fingerprint mutations, focusing on the impact of FPGA image reloads in Software Defined Radios (SDRs) when used as both transmitters and receivers. Our results highlight that FPGA reloads cause a subset of devices to exhibit two different persistent fingerprint states, one following a memoryless (Markovian) process and the other retaining temporal dependencies. Notably, we show that proper external synchronization between transmitter and receiver eliminates these fingerprint mutations, leading us to attribute the phenomenon to residual phase and timing errors rather than inherent hardware changes. Our work exposes the necessity of accounting for fingerprint dynamics caused by internal SDR events and synchronization, highlighting the limits of current measurement methodologies and the need for new, statistically robust approaches to physical-layer device identification.
We study the feasibility of black-box radio frequency fingerprint (RFF) spoofing, where an adversary lacks access to the target receiver's data or model. We present a surrogate-guided generative perturbation framework that jointly trains a generator with feedback from multiple surrogate receivers to synthesize low-power, fingerprint-level perturbations. Using real RF fingerprint datasets, we evaluate the spoofing effectiveness of forged perturbations on unseen receivers. Our results show that, even without any feedback from the target, an attacker can successfully impersonate a selected transmitter at an unseen receiver. However, the success remains limited and varies across devices, reflecting both the feasibility and the boundary of black-box RF fingerprint spoofing. These findings provide the first empirical evidence that multi-surrogate training can partially narrow the black-box gap in RF fingerprint spoofing.
With growing geopolitical interests in the maritime domain, security research of Underwater Acoustic Networks (UANs) gains momentum. One threat that has been receiving continuous attention for over two decades is the wormhole attack. However, despite having proposed numerous detection and prevention mechanisms, the theoretical feasibility of wormhole attacks in underwater scenarios has been motivated solely by the different propagation speeds of sound waves in water and electromagnetic waves in air. To the best of our knowledge, this work provides the first proof-of-concept to confirm the practical feasibility of wormhole attacks in a real UAN. Additionally, we carry out an in-depth analysis of the variables influencing the success of wormhole attacks and demonstrate how different UAN types can fall victim to wormholes depending on their configuration and deployment.
Time-of-arrival (ToA) estimation via cross-correlation is an essential building block of time-of-flight ranging. However, in narrowband systems, it is notoriously difficult to protect against
Distance-bounding protocols provide strong security guarantees, yet their secure realization depends on high-precision physical-layer timestamping of signal transmission and reception. In principle, modern commodity IEEE 802.11 chipsets fulfill these technical requirements. However, distance-bounding research and experimentation on Wi-Fi remain severely limited, as the required high-precision timestamping functionality on 802.11 hardware is tightly coupled to fixed protocol exchanges (e.g., FTM) implemented in closed-source firmware. This restricts the freedom of open research.Our work overcomes this hurdle by reverse-engineering the high-precision timestamping pipeline of the ESP32-C3 Wi-Fi stack. The analysis reveals that the ESP32-C3 hardware generates precise timestamps for ordinary OFDM frames, not only for FTM-associated ones. This unexposed capability enables precise timestamping of ordinary Wi-Fi traffic and forms the technical basis for implementing custom time-of-flight protocols on low-cost commodity devices. Drawing on this capability, we present the first open-source library extension exposing high-precision timestamping primitives on commodity IEEE 802.11 hardware.To demonstrate the practical utility of our library extension, we implement a proof-of-concept authenticated distance-bounding protocol. Our implementation showcases flexible timing, application-defined frame content, and fully open protocol logic beyond standardized exchanges. Experimental evaluation under line-of-sight conditions shows that proximity decisions with meter-level granularity are achievable. Overall, our work provides an open and auditable foundation for conducting and deploying security research based on time-of-flight on widely available IEEE 802.11 devices.
The rapid adoption of AR/VR devices, particularly in the gaming and entertainment sectors, has raised significant privacy concerns due to the need to share sensor data with platform servers to deliver immersive virtual experiences. One prominent privacy risk is the use of online tracking techniques by adversaries to profile and monitor users. While regulations such as GDPR and CCPA mandate user consent for collecting tracking cookies, adversaries are exploring alternative methods for user identification and monitoring. One such method is device fingerprinting, which relies primarily on network or browser characteristics (e.g., IP address, browser statistics) or unique device identities. Prior research has also explored the use of motion sensor imperfections for device fingerprinting. However, these approaches based on network, browser, and sensor imperfections are not persistent and can be easily mitigated through various countermeasures.In this work, we present a motion-sensor-based device fingerprinting technique TrackAR for AR/VR systems that can capture both the unique physical characteristics of the device and the user's distinctive movement patterns. We employ classical machine learning algorithms as well as state-of-the-art CNN models to implement TrackAR effectively. Our evaluation on a large motion sensor dataset (657 unique user-device combinations across 19 AR/VR devices) achieves 98% accuracy for both device identification and device-user combination identification. These results establish the efficacy and real-world applicability of our motion-sensor-based device fingerprinting method TrackAR, which poses a significant threat to user privacy in the context of AR/VR systems.
Finding Phones Fast: Low-Latency and Scalable Monitoring of Cellular Communications in Sensitive Areas
The widespread availability of cellular devices introduces new threat vectors that allow users or attackers to bypass security policies and physical barriers and bring unauthorized devices into sensitive areas. These threats can arise from user non-compliance or deliberate actions aimed at data exfiltration/infiltration via hidden devices, drones, etc. We identify a critical gap in this context: the absence of low-latency systems for high-quality and instantaneous monitoring of cellular transmissions. Such low-latency systems are crucial to allow for timely detection, decision (e.g., geofencing or localization), and disruption of unauthorized communication in sensitive areas. Operator-based monitoring systems, built for purposes such as people counting or tracking, lack real-time capability, require cooperation across multiple operators, and thus are hard to deploy. Operator-independent monitoring approaches proposed in the literature either lack low-latency capabilities or do not scale.We propose WaveTag, the first low-latency, operator-independent, and scalable system designed to monitor 5G and LTE connections across all operators prior to any user data transmission. WaveTag consists of several downlink receivers and a distributed network of uplink receivers that measure both downlink protocol information and uplink signal characteristics at multiple locations to gain a detailed spatial image of uplink signals. WaveTag then aggregates the recorded information, processes it, and provides a decision about the connection before the UE completes connection establishment. To evaluate WaveTag, we deployed it in the context of geofencing, where WaveTag was able to determine whether the signals originate from inside or outside of an area within 2.3 ms of the initial base station-to-device message, therefore enabling prompt and targeted suppression of communication before any user data was transmitted. WaveTag achieved 99.66% geofencing classification accuracy, using only the information collected before connection establishment. Finally, we conduct a real-world uplink measurement evaluation on a commercial 5G SA network.
Randomized MAC addresses aim to prevent passive device tracking, yet Wi-Fi management frames still leak structured behavioral patterns. Prior work has relied primarily on syntactic probe-request features such as Information Elements (IEs), sequence numbers (SEQ), or RSSI correlations, which degrade in dense environments and fail under aggressive randomization. We introduce StateFi, a fingerprinting framework that models device behavior as finite-state machines (FSMs), capturing both structural transition patterns and temporal execution logic. These FSMs are embedded into compact feature vectors that support efficient similarity computation and supervised classification. Across five heterogeneous campus environments, StateFi achieves 94-97% accuracy for in-network fingerprinting using full management-frame FSMs. With probe-only FSMs, it re-identifies devices under MAC randomization with up to 97% accuracy across large public datasets comprising more than a million frames. When looking at the discrimination accuracy of the model, StateFi reaches 98%, outperforming the strongest prior signature by up to 17 percentage points. These results demonstrate that FSM-level behavioral dynamics form a powerful and largely unmitigated side channel, stable enough to defeat randomization and expressive enough for robust, scalable device identification.
The security of the entire Wireless LAN (WLAN) stack heavily relies on the security of the proprietary firmware running on the radio chip. Analyzing this firmware is difficult due to its high complexity, proprietary architecture and lack of debugging capabilities. This project aims to improve the security of low-level WLAN implementations by enabling emulated fuzzing of WLAN firmware built for an obscure and proprietary architecture. This is facilitated by combining approaches from previous works on analyzing, emulating and fuzzing cellular baseband and Bluetooth firmware. After analyzing the firmware, emulation and fuzzing is performed in two steps: First, the most exposed procedures are extracted from the firmware and fuzzed separately in user mode. Then, the entire firmware is emulated and fuzzed as a whole to find more complex issues resulting from interaction of internal components. Improved security analysis for WLAN firmware helps improve the security of the firmware and, by consequence, of the entire WLAN stack.
This paper extends Antenna Subset Modulation (ASM) to the near-field regime. While far-field ASM provides angular-domain secrecy, it fails for same-angle eavesdroppers. Leveraging the quadratic phase characteristic in the near-field, we propose a weighted antenna selection strategy that enhances constellation distortion at Eve. Simulation results demonstrate significant improvements in link secrecy performance.
The 5G core network's control plane operates under a trust-based model. While this model is sufficient for single-operator deployments, it breaks down in cross-operator settings, such as Local Breakout (LBO) roaming and light MVNO-based network sharing.In this work, we revisit this trust assumption from a charging accountability perspective. We systematically identify key attack surfaces in cross-operator charging workflows by analyzing charging-relevant operations across network functions, where cross-operator visibility is inherently limited. Based on these insights, we derive design goals for accountable charging and outline a Trusted Execution Environment (TEE)-based approach that enables verifiable execution of charging-critical functions.We further develop a roaming testbed using Open5GS integrated with an online charging system, and conduct a preliminary evaluation of the TEE-based approach by measuring its overhead on charging-critical operations.
A viable remote attack surface of smartphones is the baseband chip, which handles the communication with cellular networks. One attack that is often conducted, e.g., to track users, is the deployment of Rogue Base Stations (RBSs). We built and actively maintain CellGuard, an iOS app that analyzes the interaction of an iPhone with its baseband chip to detect whether the phone connects to an
Mobile roaming ensures seamless international connectivity, but introduces additional security and performance challenges due to inter-operator communication. We present a 5G Standalone (SA) roaming testbed for evaluating the impact of Internet Protocol Security (IPSec)-based security mechanisms on system performance, built using Open5GS and PacketRusher.Our demo considers multiple cryptographic algorithms, including classical schemes and selected Post Quantum Cryptography (PQC) primitives, within the IPSec framework. We evaluate control-plane procedures, such as registration and Protocol Data Unit (PDU) session establishment times, and user-plane performance using ping latency, file download, and iPerf throughput. A Graphical User Interface (GUI) enables flexible configuration and real-time visualization. Our goal is to demonstrate the feasibility of secure IPSec deployment in practical 5G roaming scenarios.
Venue
Conference Sponsors
Travel Grants Sponsor
Corporate Sponsors
