Prevent Information Leakage Research Articles

Balancing Privacy and Robustness in Prompt Learning for Large Language Models

This paper tackles the critical issue of privacy in Natural Language Processing (NLP) systems that process sensitive data by introducing a novel framework combining differential privacy and adversarial training. The proposed solution ensures formal privacy guarantees by minimizing the influence of individual data points on the model’s behavior, effectively preventing information leakage. Simultaneously, adversarial training is applied to strengthen model robustness against privacy attacks by exposing it to adversarial examples during training. The framework is rigorously evaluated across various NLP tasks, demonstrating its capability to balance privacy preservation with high utility effectively. These results mark a significant advancement in developing secure and reliable NLP systems, particularly for applications requiring stringent data confidentiality, such as healthcare and finance.

Predicting interfacial tension in brine-hydrogen/cushion gas systems under subsurface conditions: Implications for hydrogen geo-storage

Underground hydrogen storage (UHS) critically relies on cushion gas to maintain pressure balance during injection and withdrawal cycles, prevent excessive water inflow, and expand storage capacity. Interfacial tension (IFT) between brine and hydrogen/cushion gas mixtures is a key factor affecting fluid dynamics in porous media. This study develops four machine learning models— Decision Trees (DT), Random Forests (RF), Support Vector Machines (SVM), and Multi-Layer Perceptrons (MLP)—to predict IFT under geo-storage conditions. These models incorporate variables such as pressure, temperature, molality, overall gas density, and gas composition to evaluate the impact of different cushion gases. A group-based data splitting method enhances the realism of our tests by preventing information leakage between training and testing datasets. Shapley Additive Explanations (SHAP) reveal that while the MLP model prioritizes gas composition, the RF model focuses more on operational parameters like pressure and temperature, showing distinct predictive dynamics. The MLP model excels, achieving coefficients of determination (R2) of 0.96, root mean square error (RMSE) of 2.10 mN/m, and average absolute relative deviation (AARD) of 3.25%. This robustness positions the MLP model as a reliable tool for predicting IFT values between brine and hydrogen/cushion gas (es) mixtures beyond the confines of the studied dataset. The findings of this study present a promising approach to optimizing hydrogen geo-storage through accurate predictions of IFTs, offering significant implications for the advancement of energy storage technologies.

Improving protection reliability of series‐compensated transmission lines by a fault detection method through an ML‐based model

AbstractThis article addresses the distance protection challenges associated with the series‐compensated transmission lines and the impact of fault resistance by employing a machine‐learning model. In the proposed model, stacked layers of bidirectional long short‐term memory (Bi‐LSTM) cells are fed by voltage and current signals to distinguish between different fault scenarios. This method takes advantage of only local bus measurements to prevent information leakage in communication channels. Moreover, to make the proposed method harmonics‐robust and improve the correlation interpretation between the features for the Bi‐LSTM model, the 3‐phase raw measurement signals are passed through a discrete Fourier transform (DFT) which extracts their fundamental frequency component magnitudes and angles. Then, an extensive amount of fault scenarios including different compensation levels, fault resistances, and fault locations in normal and power‐swing operational conditions are simulated to train the model. Finally, to validate the performance of the proposed protection method in the series‐compensated transmission lines, distinctive studies are also carried out based on electromagnetic transient simulations. The obtained results confirm the remarkable performance of the proposed method in discriminating fault types, faulty phases, internal or external faults, and normal or power‐swing conditions of the power system.

“Will artificial intelligence platforms replace designers in the future?” analyzing the impact of artificial intelligence platforms on the engineering design industry through color perception

This research investigates the impact of artificial intelligence platforms on the engineering design industry by analyzing the perceptions of color and views on artificial intelligence development among designers and non-designers. The study utilized a two-stage approach: the first involved artificial intelligence-assisted design tasks and focus group discussions with 24 participants, capturing both visual and textual data. K-means color clustering and thematic analysis were applied to images and texts. The second stage involved surveying 222 individuals and conducting a chi-square test to analyze the frequency of different color dimensions used by the two groups. The results demonstrate that differences in color perception influenced design decision-making outcomes. There was widespread acknowledgment across all groups of the efficiency and inspirational potential of artificial intelligence platforms. However, variations in design education resulted in differing opinions on the replaceability of designers. Future trajectories for artificial intelligence platforms are likely to focus on specialization while addressing challenges such as echo chamber effects, copyright disputes, and the prevention of private information leakage.

Stretchable MWCNTs-OH/PDMS composite elastomer with hierarchical porous structure for wideband THz absorption

It is important to develop a wideband THz absorber for the prevention of terahertz electromagnetic pollution and information leakage. Some commonly used methods, such as metamaterials or dynamic modulation technology, have played important roles in the study of wideband THz absorbers. However, most of these absorbers are rigid or non-stretchable, which limits the practical applications in large mechanical deformation and non-plane scenarios. In the paper, we proposed a stretchable MWCNTs-OH/PDMS composite elastomer using the sugar template method. A series of characterization methods were carried out to verify the excellent compatibility of PDMS and MWCNTs-OH. Benefiting from its hierarchical porous structure, the draw-length ratio of sample and tensile strength reaches ∼ 190% and ∼ 0.4 MPa, respectively. More THz waves could be trapped inside the sample for conductive loss because of better impedance matching and conductivity loss (the smallest square value is ∼0.4 kΩ/□), which leads to wideband and high absorptivity (over 98% in 0.2∼2.0 THz range). Furthermore, the tested value of EMI SE was over 10 dB in 0.22∼1.82 THz frequency range at transmission mode. Combined with inherent hydrophobicity characteristics, the material also could be used in humid and rainy environments. The systemic study of stretchable MWCNTs-OH/PDMS composite elastomer will provide theoretical and technical support for subsequent THz absorption in practical scenarios.

Battery recycling and coordination in information leakage prevention under blockchain technology in a new energy vehicles supply chain

Retired battery recycling and information leakage prevention under blockchain introduction intertwine in the new energy vehicle (NEVs) supply chain, but rare literature has jointly investigated them. Therefore, we characterize the vehicle goodwill under recycling and marketing efforts as well as the automobile demand function with and without blockchain introduction, identify the long-run discounted profits of the manufacturer and retailers in the supply chain, and investigate the Stackelberg differential game equilibrium under four scenarios: no blockchain introduction, manufacturer introduction, manufacturer-retailer cooperative introduction, and cooperative introduction plus wholesale price discounts for preventing information leakage. The results show that the manufacturer can introduce blockchain when costs are below a certain threshold, which increases the recycling rate of decommissioned batteries and improves goodwill, consumer demand, and marginal profits for the manufacturer and retailer; the retailer participation in the introduction of blockchain solves the free-rider problem and reduces the thresholds introduced, but does not affect battery recycling, goodwill, consumer demand, and wholesale and retail prices; wholesale price discounts effectively discourage the retailer from leaking information about the manufacturer and lead to improved profits for the retailer, but results in higher earnings for the manufacturer only when discounts are below a certain threshold.

Anti‐leakage transmission method of high privacy information in electric power communication network based on digital watermarking technology

AbstractA method of transmission of highly confidential information in power communication network based on digital watermarking technology is proposed in order to reduce the leakage risk of power information and achieve the purpose of safe transmission of private information. This method establishes the data topology model of power communication network and connects the data receiving and sending terminals. Digital watermarking technology is used to embed digital watermarking in the information to be transmitted, encrypt the private information, and optimize the DES encryption algorithm to encrypt the information twice, so as to realize the safe transmission of information. The experimental results show that the digital watermark embedding rate, DES encryption rate, DES decryption rate and watermark extraction rate of the proposed method are all above 90 Mbps, and the file transfer time of 3965Byte is less than 5s, leakage risk rate and packet loss rate are 0.0001% and 0.006%, respectively, which effectively protects the security of high privacy information in the power communication network and prevents information leakage.

A Stealthy Communication Model with Blockchain Smart Contract for Bidding Systems

With the widespread adoption of blockchain technology, its public ledger characteristic enhances transaction transparency but also amplifies the risk of privacy breaches. Attackers can infer users’ real identities and behaviors by analyzing public transaction patterns and address relationships, posing a severe threat to users’ privacy and security, and thus hindering further advancements in blockchain applications. To address this challenge, covert communication has emerged as an effective strategy for safeguarding the privacy of blockchain users and preventing information leakage. But existing blockchain-based covert communication schemes rely solely on the immutability of blockchain itself for robustness and suffer from low transmission efficiency. To tackle these issues, this paper proposes a stealthy communication model with blockchain smart contract for bidding systems. The model initiates by preprocessing sensitive information using a secret-sharing algorithm-the Shamir (t, n) threshold scheme-and subsequently embeds this information into bidding amounts, facilitating the covert transfer of sensitive data. We implemented and deployed this model on the Ethereum platform and conducted comprehensive performance evaluations. To assess the stealthiness of our approach, we employed a suite of statistical tests including the CDF, the Kolmogorov–Smirnov test, Welch’s t-test and K–L divergence. These analyses confirmed that amounts carrying concealed information were statistically indistinguishable from regular transactions, thus validating the effectiveness of our solution in maintaining the anonymity and confidentiality of information transmission within the blockchain ecosystem.

Exploring Zero-Day Attacks on Machine Learning and Deep Learning Algorithms

In the rapidly evolving field of artificial intelligence, machine learning (ML) and deep learning (DL) algorithms have emerged as powerful tools for solving complex problems in various domains, including cyber security. However, as these algorithms become increasingly prevalent, they also face new security challenges. One of the most significant of these challenges is the threat of zero-day attacks, which exploit unknown and unpredictable vulnerabilities in the algorithms or the data they process. This paper provides a comprehensive overview of zero-day attacks on ML/DL algorithms, exploring their types, causes, effects, and potential countermeasures. The paper begins by introducing the concept and definition of zero-day attacks, providing a clear understanding of this emerging threat. It then reviews the existing research on zero-day attacks on ML/DL algorithms, focusing on three main categories: data poisoning attacks, adversarial input attacks, and model stealing attacks. Each of these attack types poses unique challenges and requires specific countermeasures. The paper also discusses the potential impacts and risks of these attacks on various application domains. For instance, in facial expression recognition, an adversarial input attack could lead to misclassification of emotions, with serious implications for user experience and system integrity. In object classification, a data poisoning attack could cause the algorithm to misidentify critical objects, potentially endangering human lives in applications like autonomous driving. In satellite intersection recognition, a model stealing attack could compromise national security by revealing sensitive information. Finally, the paper presents some possible protection methods against zero-day attacks on ML/DL algorithms. These include anomaly detection techniques to identify unusual patterns in the data or the algorithm’s behaviour, model verification and validation methods to ensure the algorithm’s correctness and robustness, federated learning approaches to protect the privacy of the training data, and differential privacy techniques to add noise to the data or the algorithm’s outputs to prevent information leakage. The paper concludes by highlighting some open issues and future directions for research in this area, emphasizing the need for ongoing efforts to secure ML/DL algorithms against zero-day attacks.

Multidimension‐Regulated Dynamic Timing Control Over Multicolor Emission of Host–Guest Assemblies for Information Encryption and Advanced Anti‐Counterfeiting

AbstractDynamical control over molecular luminescence, especially in a time‐dependent manner, holds great promise for the development of smart luminescent materials for anti‐counterfeiting and preventing information leakage. Herein, a series of self‐assembled systems are reported using pillar[5]arene (DMP[5]) and spiropyran derivatives (SP‐C4‐Py). The assemblies rely on the time‐encoded locking and unlocking ring‐switching and fluorescence resonance energy transfer (FRET) units for information camouflage and multilevel encryption. DMP[5] with cyan solid fluorescence color acts as the host and energy transfer (EnT) donor, while photochromic SP‐C4‐Py with the ring‐opened and closed isomers acts as guest and EnT acceptor. When irradiated, the assemblies undergo a time‐dependent luminescence color change ranging from cyan to yellow to red through a FRET process. The molar ratio of host and guest in the assembly systems affects the FRET efficiency, and the power of the irradiation source influences the isomerization degree and rate of SP‐C4‐Py, allowing for precise control over the fluorescent color transition time. The combination of molecular composition and external stimuli governs the kinetics of color change, resulting in a difference in the appearance time of a specific fluorescent color pre‐designed as correct authorized information. By combining these diverse assembles in one label, information encryption and dynamic information identification are achieved in the dimensions of time, ratio, and light power. This time‐dependent feature offers the assembly materials with a multilevel security and provides new possibilities for anti‐counterfeiting and blocking information leakage.

VCGERG

Vulnerability can lead to data loss, privacy leakage and financial loss. Accurate detection and identification of vulnerabilities is essential to prevent information leakage and APT attacks. This paper explores the possibility of digging the valuable information in vulnerability reports deeply. We propose a new model, VCGERG, which products a graph using key information from vulnerability reports and embeds the graph into the vector space using a keywords-LINE graph embedding algorithm based on the attention of neighboring nodes. VCGERG model uses the OVR random forest algorithm to classify vulnerabilities. Our model can get the complicated local and global information of the graph in large-scale dataset and achieve better results. In order to verify the effectiveness of our model, it is evaluated on many experiments. Compared with other models, our method has a higher accuracy rate of 0.975.

Precursor of privacy leakage detection for individual user

The pursuit of dividends from the burgeoning realm of data traffic has emerged as a prevailing trend. Given this current state of affairs, the safeguarding of personal information has become an urgent task. Current methods for personal privacy protection primarily offer alerts when information breaches occur, but at that stage, irreversible breaches have already transpired. Thus, the study of preemptive privacy breach prediction is a task of significant importance within the realm of privacy protection. However, the endeavor to predict privacy breaches in advance remains exceedingly challenging, owing to several factors: (i) The complexity of social networks gives rise to high-dimensional features. (ii) Concerning the comprehensive and precise capture of pathways leading to information leakage. (iii) How to employ time series data effectively to realize early predictions.This study proposes a novel approach for constructing graph neural networks and forecasting privacy breaches, centered on the context of user-generated content within specific time frames, integrates spatial graph structures with temporal series information. The integration can entirely achieve the advance prediction of users' privacy status, thereby preventing information leakage. Empirical tests on real-world datasets demonstrate that our approach surpasses traditional time series forecasting methods in privacy breach predictions, achieving a notable average improvement of 2% in F1 score, Recall, and Precision metrics.

Computer intelligent network security and preventive measures of internet of things devices

AbstractThe paper focused on researching and analyzing computer intelligence network security and preventive measures in the context of the IoT, aiming to improve the security coefficient of the IoT network and reduce IoT network security accidents through computer intelligence technology. Through experiments, we obtained data that demonstrated the effectiveness of computer intelligence in improving IoT security. In several groups of experiments, the maximum number of information leaks in the IoT network using computer intelligence within a month was 10 times smaller than the maximum number in traditional IoT networks, and the minimum number was 8 times smaller. This shows that computer intelligence can prevent information leakage in the IoT. Similarly, in several groups of experiments, the maximum number of data thefts in a month in the IoT network using computer intelligence was 15 times smaller than the maximum number in traditional IoT networks, and the minimum number was 16 times smaller. This demonstrates that computer intelligence can prevent data theft in the IoT. These findings confirm that computer intelligence can improve the security of the IoT network.

Towards trustworthy and privacy-preserving decentralized auctions

Blockchain smart-contracts can be used as service mappers, connecting a contractor with the service provider best fitting desired service requirements (e.g., price or quality of service). The allocation consists of comparing competitive bids using a smart-contract. However, in competitive environments, service providers may be reluctant to share sensitive information offers with the blockchain as it makes any transaction implicitly public. To reconcile data privacy imperatives with the benefits of blockchain, we propose to leverage fully homomorphic encryption (FHE) for blockchain-based sealed-bid auctions. More precisely (i) FHE enables the processing of bids without decrypting them, (ii) smart-contracts gather and orchestrate bids comparison, and (iii) a computation oracle carries on comparisons over ciphered data. Collusion attempts may occur between bidders and the computation oracle. To prevent this, we combine FHE with hybrid RSA/AES encryption to preserve the privacy of the onchain bid contents. Hence, our protocol prevents information leakage onchain and on the service providers’ side during bids comparison. We validate this approach through an implemented prototype.

PPSFL: Privacy-Preserving Split Federated Learning for heterogeneous data in edge-based Internet of Things

With the rapid increase in the number of Internet of Things (IoT) devices and the amount of data they generate, the traditional cloud-based approach is gradually unable to meet the actual needs of many scenarios. Distributed collaborative machine learning (DCML) paradigms such as Federated Learning (FL) and Split Learning (SL) provide possibilities for effective use of decentralized data in edge-based IoT. However, critical challenges in terms of data privacy, heterogeneity, and constrained resources remain to be handled. Despite extensive efforts, current solutions still cannot address the above challenges simultaneously. Therefore, studies in this emerging research field remain inadequate. In this paper, we propose a hybrid framework for combining FL with SL, named Privacy-Preserving Split Federated Learning (PPSFL). It facilitates privacy protection with an appropriate model decomposition strategy and mitigates the negative impact of data heterogeneity by incorporating private Group Normalization (GN) layers into the network. Extensive empirical results demonstrate that PPSFL attains better performance than other state-of-the-art distributed collaborative learning methods on different datasets. We also evaluate and compare the resistance of all baselines to reconstruction attacks with various image datasets. Results supported by comparative experiments indicate that our method can greatly prevent information leakage from raw data while maintaining classification performance. Additionally, the comparisons in terms of communication and computation overhead show that PPSFL is also competitive.

Federated Unlearning and Its Privacy Threats

Federated unlearning has emerged very recently as an attempt to realize "the right to be forgotten" in the context of federated learning. While the current literature is making efforts on designing efficient retraining or approximate unlearning approaches, they ignore the information leakage risks brought by the discrepancy between the models before and after unlearning. In this paper, we perform a comprehensive review of prior studies on federated unlearning and privacy leakage from model updating. We propose new taxonomies to categorize and summarize the state-of-the-art federated unlearning algorithms. We present our findings on the inherent vulnerability to inference attacks of the federated unlearning paradigm and summarize defense techniques with the potential of preventing information leakage. Finally, we suggest a privacy preserving federated unlearning framework with promising research directions to facilitate further studies as future work.

Protecting FPGA-Based Cryptohardware Implementations from Fault Attacks Using ADCs.

The majority of data exchanged between connected devices are confidential and must be protected against unauthorized access. To ensure data protection, so-called cryptographic algorithms are used. These algorithms have proven to be mathematically secure against brute force due to the key length, but their physical implementations are vulnerable against physical attacks. The physical implementation of these algorithms can result in the disclosure of information that can be used to access confidential data. Some of the most powerful hardware attacks presented in the literature are called fault injection attacks. These attacks involve introducing a malfunction into the normal operation of the device and then analyzing the data obtained by comparing them with the expected behavior. Some of the most common methods for injecting faults are the variation of the supply voltage and temperature or the injection of electromagnetic pulses. In this paper, a hardware design methodology using analog-to-digital converters (ADCs) is presented to detect attacks on cryptocircuits and prevent information leakage during fault injection attacks. To assess the effectiveness of the proposed design approach, FPGA-based ADC modules were designed that detect changes in temperature and supply voltage. Two setups were implemented to test the scheme against voltage and temperature variations and injections of electromagnetic pulses. The results obtained demonstrate that, in 100% of the cases, when the correct operating voltage and temperature range were established, the detectors could activate an alarm signal when the cryptographic module was attacked, thus avoiding confidential information leakage and protecting data from being exploited.

A quantum blind signature scheme based on dense coding for non-entangled states

In some schemes, quantum blind signatures require the use of difficult-to-prepare multiparticle entangled states. By considering the communication overhead, quantum operation complexity, verification efficiency and other relevant factors in practical situations, this article proposes a non-entangled quantum blind signature scheme based on dense encoding. The information owner utilizes dense encoding and hash functions to blind the information while reducing the use of quantum resources. After receiving particles, the signer encrypts the message using a one-way function and performs a Hadamard gate operation on the selected single photon to generate the signature. Then the verifier performs a Hadamard gate inverse operation on the signature and combines it with the encoding rules to restore the message and complete the verification. Compared with some typical quantum blind signature protocols, this protocol has strong blindness in privacy protection, and higher flexibility in scalability and application. The signer can adjust the signature operation according to the actual situation, which greatly simplifies the complexity of the signature. By simultaneously utilizing the secondary distribution and rearrangement of non-entangled quantum states, a non-entangled quantum state representation of three bits of classical information is achieved, reducing the use of a large amount of quantum resources and lowering implementation costs. This improves both signature verification efficiency and communication efficiency while, at the same time, this scheme meets the requirements of unforgeability, non-repudiation, and prevention of information leakage.

Data encoding for healthcare data democratization and information leakage prevention

The lack of data democratization and information leakage from trained models hinder the development and acceptance of robust deep learning-based healthcare solutions. This paper argues that irreversible data encoding can provide an effective solution to achieve data democratization without violating the privacy constraints imposed on healthcare data and clinical models. An ideal encoding framework transforms the data into a new space where it is imperceptible to a manual or computational inspection. However, encoded data should preserve the semantics of the original data such that deep learning models can be trained effectively. This paper hypothesizes the characteristics of the desired encoding framework and then exploits random projections and random quantum encoding to realize this framework for dense and longitudinal or time-series data. Experimental evaluation highlights that models trained on encoded time-series data effectively uphold the information bottleneck principle and hence, exhibit lesser information leakage from trained models.

Computer Network Virus and Computer Network Security Prevention

With the rapid development of information technology, network security management is becoming increasingly complex, especially, from the point of view of data security, the protection of data should be strengthened, now, the popularity of computers and smart phones will bring a large amount of information, which is inextricably linked with our property and personal privacy, in the human society, with the continuous development of intelligent technology, people's personal information and property protection are also paying more and more attention. Therefore, in order to achieve the purpose of network security and preventing information leakage, it is necessary to analyze and categorize all kinds of data and build a unified management platform on this basis. In recent years, the development of China's computer technology has been more and more attention, especially with the rapid development of big data, data storage, processing capacity, security and other aspects of the rapid development of all walks of life has brought great economic benefits, and the development of social productivity has played a huge role in promoting, but also brought a huge security risks, but the combination of big data technology and computer network security is a good technology, and its advantages are fully reflected in daily life.