Smart grids enable advanced functions like power dispatching, real-time pricing, and electricity theft detection. However, these capabilities have also raised significant concerns about user privacy. Privacy-preserving data aggregation technology enables the collection and utilization of users’ power consumption data without compromising their privacy. However, existing multidimensional data aggregation approaches fail to ensure user privacy and provide query services for data consumers in different roles. To address these problems, a multidimensional data aggregation scheme with function query (MDA-FQ) is proposed to realize the value of the multidimensional consumption data while safeguarding user privacy. First, MDA-FQ employs inner-product functional encryption and the Chinese Remainder Theorem to securely encrypt users’ multidimensional power consumption data. Second, a role-based data access mechanism is introduced to enable data consumers with varying permissions to request different types of data from cloud service providers, thus maximizing the utility of multidimensional data under privacy constraints. Third, MDA-FQ incorporates differential privacy on multidimensional data to ensure that each data dimension is protected against differential attacks, thereby enhancing overall privacy protection. Security analyses indicate that MDA-FQ can effectively counter differential attacks and other privacy risks. Extensive experimental evaluations further demonstrate that MDA-FQ not only supports versatile queries on multidimensional data but also significantly reduces computation and communication overhead, making it both practical and scalable.
Physical-layer authentication (PLA) provides an important approach to achieve secure access by exploiting the inherent hardware characteristics of devices. Radio frequency (RF) fingerprinting-based PLA emerges as a cost-effective solution to counter identity spoofing threats. Although deep learning (DL) has been widely adopted for RF fingerprint extraction and device identification, existing methods typically require complex DL models with massive parameters to achieve high accuracy, making applying them to resource-constrained devices impractical. Furthermore, most approaches focus on closed-set recognition scenarios, failing to address practical environments containing unknown devices. To overcome these limitations, we propose LRF-KDTL, a novel lightweight DL-based RF fingerprint identification method leveraging knowledge distillation and transfer learning. Specifically, LRF-KDTL employs a teacher-student framework for feature distillation and subsequently retrains the student model through transfer learning, achieving precise device identification with significantly reduced complexity. Extensive experiments in open-set scenarios demonstrate that the proposed method reaches 90.21
Recently, Karmakar et al. designed a mutual authentication and key agreement mechanism for an unmanned aerial vehicle (UAV) and the ground station. They claimed that their mechanism is resilient against common security threats. However, our analysis reveals that their mechanism fails to resist eavesdropping attack and ephemeral secret leakage attack. Moreover, this mechanism further becomes vulnerable to the impersonation attack when the ephemeral secret is leaked. Specifically, an adversary can obtain the session key established in the current session in the event that an ephemeral secret is leaked. After identifying the root causes of these vulnerabilities, we propose targeted countermeasures to strengthen protocol security.
The paper presents the microeconomic theory of marginal utility altered according to the terms of cybersecurity investment in healthcare. As part of the paper, cooperation was initiated with a Czech hospital which unsuccessfully applied for a subsidy from the National Recovery Plan. Due to the rejection of the subsidy, but the still underdeveloped state of cybersecurity in the healthcare facility, the author was approached with a request for assistance in negotiating financial support with the hospital’s top management from its own resources. The paper alters the ROSI model of return on security investment in order to clearly show the impact of the proposed investments on each protected person, and especially their marginal financial burden on the entire facility. The presented alteration of the ROSI model is applicable across the healthcare sector and beyond, especially in the industries that face difficulties counting the exact cost of data loss.
With the rapid advancement of network technology, incidents of network attacks are increasingly frequent, and malware has become a commonly used tool in such attacks. Dynamic analysis plays a crucial role for security professionals in understanding and assessing malware, allowing for the capture of dynamic behaviors within a sandbox environment. However, determining the maliciousness of these behaviors often requires substantial expert knowledge, making the process both time-consuming and challenging. To address this, we propose an automated technique for determining sandbox analysis reports as malicious or benign. Our approach combines dynamic features extracted from sandbox reports with a large language model. By fine-tuning a model based on Llama3, we achieve accurate determination between malicious and benign samples, reaching a precision of 96.82
Binary Code Similarity Detection (BCSD) has become a pivotal area of research in cybersecurity, addressing critical applications such as vulnerability detection, malware analysis, and software plagiarism. This review synthesizes recent advancements in BCSD, highlighting the integration of machine learning techniques, the importance of semantic understanding, and the incorporation of domain knowledge. Key challenges, including semantic equivalence, representation of control flow, and scalability, are discussed alongside innovative methodologies that enhance detection accuracy. The review also explores the implications of recent studies on practical applications, including vulnerability search systems and malicious code detection, while identifying future research directions aimed at improving model interpretability, standardizing benchmarks, and addressing the complexities of diverse hardware architectures.
With the widespread use of Internet of Things (IoT) devices, hidden Web interfaces have become one of the core security threats. These interfaces often lack authentication, making them prime targets for attackers to steal sensitive information or manipulate devices. Existing solutions only analyze the security of publicly exposed Web interfaces and cannot effectively identify deeply embedded private interfaces. To address these challenges, we propose HWISeeker, a novel solution that combines static analysis and semantic reasoning techniques to discover and verify hidden Web interfaces in IoT devices automatically. Specifically, HWISeeker filters the key keywords associated with the interfaces through heuristic rules and semantic analysis, employs a control flow backtracking algorithm to restore the logical relationships of the keywords, and finally assembles and validates the legitimacy of the interfaces using black-box fuzzing. HWISeeker was evaluated on nine commercial IoT devices and successfully identified 175 hidden Web interfaces, achieving 2.8 times the detection capability of the current state-of-the-art solution, IoTScope. Furthermore, 23 vulnerabilities were detected, 18 of which have been assigned CVE IDs.
Dynamic pricing is a hot topic in the data application-driven industry and academia in the emerging data security market. Traditional pricing methods cannot capture the value changes in the time dimension and cannot be flexibly adjusted according to market changes. The Deep Q-Network (DQN) in reinforcement learning also has an overestimation bias in the target value calculation process. To address the above challenges, we propose a dynamic pricing model combined with a reinforcement learning framework. This model can cleverly apply the time perspective of data streams to the actual data market. Firstly, we design models including static and dynamic price profit maximization. Then, we propose a new dynamic pricing method by combining the Double Deep Q-Network (DDQN) with the Noisy Network (Noisy Net) to monitor market environment changes in real time and improve the overall efficiency of the data market. To verify the advantages and effectiveness of this method, we generate consumer sample data additionally for experimental observation and compare the capabilities of static and dynamic pricing methods. A large number of comparative experimental results demonstrate the effectiveness of our proposed method and its excellent performance in dynamic pricing scenarios.
Double-base chain (DBC) has been extensively studied for speeding up scalar multiplication, which significantly impacts the execution efficiency of ECC. In this paper, we estimate the theoretically optimal upper bound and lower bound for the leading term of a canonic DBC. Firstly, we develop a new method to estimate the bounds and prove its correctness, based on Yu et al.’s work in EUROCRYPT 2020. Secondly, we obtain the new bounds by applying this new method. Thirdly, we compare the efficiency improvements of canonic DBC generation algorithms and scalar multiplication algorithms using the new theoretically optimal upper bound. Compared to NAF algorithm, our new scalar multiplication algorithm achieves about 13% efficiency gains. Finally, we analyze the bounds for the leading term proposed by Eom et al., identify their limitations and provide counterexamples to demonstrate the issues.
With the escalating complexity of digital infrastructure, traditional cybersecurity defense methods are increasingly limited, highlighting an urgent need for innovative attack surface generation. This study proposes Intell-dragonfly, an advanced cybersecurity attack surface generation engine leveraging cutting-edge Large Language Models (LLMs), specifically harnessing the capabilities of DeepSeek. To navigate dynamic threats and ethical considerations, Intell-dragonfly incorporates robust prompt engineering strategies, enabling the creation of diversified, personalized, and context-aware attack scenarios. Its effectiveness is quantitatively verified through rigorous experiments, demonstrating significant improvements in attack surface coverage, attack vector diversity, and generation efficiency compared to traditional methods. Furthermore, the paper provides initial empirical analysis of generated vulnerability effectiveness, showcasing its potential to assist researchers in identifying novel attack paths. Beyond this, Intell-dragonfly’s design considers enhancing interpretability and traceability, and its future evolution includes multi-agent collaborative systems. This work crucially emphasizes its application in strengthening digital infrastructure security, targeting critical systems like ICS, SCADA, and cloud environments, thus contributing to a more proactive and adaptable paradigm for automated cybersecurity defense and offensive research.
Large language models, such as ChatGPT developed by OpenAI, have significantly enhanced productivity in various software development tasks. Despite these advancements, the security of code generated by those models remains a concern. This study investigates the security of ChatGPT-generated cryptographic API code snippets in four popular programming languages: Java, Python, C, and JavaScript. Our findings reveal that ChatGPT, on average, produces 1.38
With the widespread adoption of federated learning in vehicular ad-hoc networks (VANETs), gradient privacy leakage and Byzantine attacks have emerged as critical research challenges. To address these issues, this paper proposes a privacy-preserving and Byzantine-resilient federated learning scheme in VANETs. Specifically, we use obfuscation factors to obfuscate the local gradients to ensure their privacy is not compromised. Then, we propose a method that calculates vehicles’ credibility using local gradient deviation, and this method identifies Byzantine nodes by evaluating the degree of gradient deviation in VANETs. Security analysis and experimental results show that the proposed scheme can protect the privacy of vehicles and also identify Byzantine nodes to improve the accuracy of the global model.
With the continuous development deepening of the application of location-based services, more and more attention has been paid to the privacy and security protection of location data. In recent years, differential privacy as a rigorous and provable privacy protection technology, has been more and more widely used in location privacy protection, and local differential privacy has attracted special attention due to its independence from trusted data centers. However, the existing location data publishing methods based on local differential privacy still have the following problems, such as the lack of flexible personalized privacy setting method, the excessive noise, and low computational efficiency. In response to these issues, this article proposes a personalized method based on local differential privacy and fuzzy set, which divides the spatial grid according to the privacy needs of users in different locations, limits the candidate set of user publishing locations to the adjacent spatial grid, and perturbs the membership degree on the basis of the fuzzy set. Theoretical analysis and experimental results show that the proposed method has good efficiency and data availability while satisfying the requirements of differential privacy protection effect and flexible personalized privacy setting.
Steganography is an important branch of information hiding used for covert communication. In modern steganography, steganographic coding is the key technology for content-adaptive message embedding, achieving near-optimal security performance by minimizing distortion. So far, Syndrome-Trellis Codes (STC) and Steganographic Polar Codes (SPC) have demonstrated near-optimal performance in content-adaptive steganography. However, when the cover is short, both of these codes show a significant performance gap from the rate-distortion bound. To address this limitation, we propose Steganographic Polarization-Adjusted Convolutional Codes (SPACC), a novel near-optimal steganographic coding method tailored for short cover scenarios. SPACC innovatively integrates Polarization-Adjusted Convolutional (PAC) codes, which have shown better performance than traditional polar codes in short block-length communication scenarios. This advantage motivates us to adopt PAC codes in steganographic coding. Specifically, SPACC uses the polarization channel metrics calculated by Bhattacharyya parameters to select the parity-check matrix for steganographic coding. Then, secret messages are embedded using the Successive Cancellation List (SCL) decoder of PAC codes while minimizing the total distortion. Experimental results demonstrate that our proposed SPACC achieves performance much closer to the optimal coding for short cover. Even for longer cover, it outperforms SPC in coding efficiency.
Recently, and have been proposed as new isogeny-based key exchange protocols to resist torsion point attacks. In this work, we propose two novel types of isogeny-based key exchange protocols based on and . Firstly, we propose ^ and ^ using different forms of primes in and . Secondly, we introduce a similar hardness assumption that reveals an image of torsion point of small order based on the assumption of . We show that this assumption is sufficient to build parallel isogenies and construct two key exchange protocols called ^ and ^ . Finally, we provide the corresponding parameters for the new protocol proposed at different security levels and develop a Sagemath implementation of these protocols. For example, compared to the original , when the security strength is 128 bits, ^ and ^ achieve speeds that are 2.2 × and 28.5 × faster in the key generation phase and 2.0 × and 19.1 × faster in the key exchange phase. The size of the public key is reduced by factors of 1.1 × and 2.7 × , respectively.
Enterprises today manage digital identities that span domains, social-media handles, and other unique online namespaces. This paper analyzes the main risks and opportunities: lapses in the domain-name life-cycle that invite phishing and brand impersonation; a secondary market where investors drive up costs or facilitate hostile takeovers; and misuse of corporate handles on platforms like X, Facebook, and Instagram. It contrasts Western and Chinese practices - numeric domains, new-gTLD booms, and differing regulations - and outlines a defense-in-depth strategy of timely renewals, defensive registrations, continuous impersonation monitoring, and recourse to legal remedies. Finally, it surveys emerging disruptors such as blockchain naming, further gTLD expansion, and evolving verification standards. A holistic, adaptive approach is essential to preserve brand trust worldwide.
Deep learning-based side-channel analysis has been proven to be a powerful method for attacking cryptographic devices. However, it faces a portability problem due to differences between the profiling device and the target device. At DAC 2022, Cao et al. proposed an adversarial learning-based profiling side-channel analysis (AL-PA) approach to mitigate this issue. Nevertheless, this method has not been systematically analyzed under challenging scenarios such as low signal-to-noise ratios and protection countermeasures. To address this gap, this paper investigates the performance limitations of AL-PA under challenging scenarios. We identify that fully sharing encoder parameters across domains restricts the model’s adaptability when the source and target distributions diverge significantly. To overcome this limitation, we propose a training strategy that freezes the source encoder parameters, thereby improving the stability and adaptability of adversarial training. Furthermore, a randomized multilinear mapping is introduced to replace the original conditional outer product, which reduces the number of discriminator parameters while preserving classification performance. We validate the proposed method on the SAKURA-AES and XMEGA-SM4 datasets. Experimental results show that our approach reduces discriminator parameters by at least 96.87
The imminent development in the field of quantum computing threat has the potential to undermine current cryptographic protections, enabling new forms of cyberattack (e.g., ’store now, decrypt later’ strategies) and magnifying liability issues for governments, companies, and individuals. The paper examines how existing criminal and civil liability frameworks in the European Union, United States, and China address cybercrime in general and assesses their adequacy in a post-quantum environment. Drawing on legal sources (penal codes, directives, conventions, statutes) and scholarly commentary, this paper compares the substantive offenses (unauthorized access, data interference, misuse of devices, etc.), enforcement mechanisms, and tort-like duties in each legal system. The paper considers the role of encryption-related regulation (export controls, mandatory standards), and corporate accountability rules. The conducted research highlights both commonalities (for example, criminalization of hacking and malware in all systems) and divergences (such as the emphasis of the EU on cross-border cooperation versus the approach of China’s cyber sovereignty). Although entities taken into account recognize the core cyber offenses, few legal provisions explicitly contemplate the quantum threat or impose duties to adopt quantum-safe measures. In conclusion, a state of preparedness for the coming era is considered along with recommendations for areas needing reform, such as new statutory duties to upgrade cryptography and enhanced international cooperation.
Digital systems are ubiquitous and play a pivotal role in our daily lives. However, the digital hyper-infrastructure is not without its flaws, and could be considered defective. Society-wide questions of IT security and its regulation define the field of cybersecurity. However, the term itself, as well as many of the underlying principles it encompasses, are the subject of contestation and are marked by both explicit and implicit contradictions. To illustrate this, I conduct an extensive literature review in which I analyze the scopes, actors, priorities, threats and approaches being negotiated in the cybersecurity discourse. The subsequent analysis involves the categorization of the findings into four distinct categories of cybersecurity understandings: individual IT security, civic infrastructure IT security, national IT security and holistic IT security. I then delineate the relevant differentiating dimensions, explicate the hidden tensions and discuss the potential of data protection theory to inform a more systematic cybersecurity discourse. I argue that it is imperative that not only scholars explicitly accept the multiplicity of cybersecurities if they are to facilitate fruitful discourse and provide actionable insights for politics, regulation and practice towards securing the digital hyper-infrastructure.
Cyber attacks, including cyber social engineering attacks, such as malicious emails, are always evolving with time. Thus, it is important to understand their evolution. In this paper we characterize the evolution of malicious emails through the lens of Psychological Factors (PFs), which are humans’ psychological attributes that can be exploited by malicious emails (i.e., attackers who send them). For this purpose, we propose a methodology and apply it to conduct a case study on 1,260 malicious emails over a span of 21 years (2004–2024). Our findings include: (i) attackers have been constantly seeking to exploit many PFs, especially the ones that reflect human traits; (ii) attackers have been increasingly exploiting 9 PFs and mostly in an implicit or stealthy fashion; (iii) some PFs are often exploited together. These insights shed light on how to design future defenses against malicious emails.
