Subgraph matching plays an important role in electronic design automation (EDA) and circuit verification. Traditional rule-based methods have limitations in generalizing to arbitrary target circuits. Furthermore, node-to-node matching approaches tend to be computationally inefficient, particularly for large-scale circuits. Deep learning methods have emerged as a potential solution to address these challenges, but existing models fail to efficiently capture global subgraph embeddings or rely on inefficient matching matrices, which limits their effectiveness for large circuits. In this paper, we propose an efficient graph matching approach that utilizes Graph Neural Networks (GNNs) to predict regions of high probability for containing the target circuit. Specifically, we construct various negative samples to enable GNNs to accurately learn the presence of target circuits and develop an approach to directly extracting subgraph embeddings from the entire circuit, which captures global subgraph information and addresses the inefficiency of applying GNNs to all candidate subgraphs. Extensive experiments demonstrate that our approach significantly outperforms existing methods in terms of time efficiency and target region prediction, offering a scalable and effective solution for subgraph matching in large-scale circuits. The source code is available at https://github.com/sang-woo-seo/Circuit-matching-GNN.

Odem softwarization, where baseband signals are fully processed using software on a general-purpose CPU, is a promising technology in mobile communications due to its simplicity and flexibility in realizing various features. On the other hand, many still question the effectiveness of a softwarized modem in commercial environments concerning performance and complexity. Motivated by this perspective, this paper presents the design and implementation of a softwarized modem with the specific feature of 5G cell search for field evaluation. Based on the baseline algorithms of 5G cell search in Open Air Interface (OAI), we propose a new software architecture which can efficiently manage a 5G cell search procedure and decompose the overall 5G cell search into sub-algorithms. We also design and implement novel sub-algorithms that enhance the detection of Synchronization Signal Blocks (SSBs). Our softwarized modem utilizes dual-rate sampling to significantly reduce computation complexity during timing offset estimation. It also adaptively detects synchronization signals or cell identities based on the presence of inter-cell interference or multi-path fading. The performance evaluation through field experiments concludes that our softwarized modem outperforms the baseline, and the proposed sub-algorithms are effective in enhancing cell search performance. The detection probability and time consumption results for our softwarized modem confirm that it is feasible for commercial uses.M

This paper investigates the use of through-skull sound conduction to authenticate smartglass users. We mount a surface transducer on the right mastoid process to play cue signals and capture skull-transformed audio responses through contact microphones on various skull locations. We use the resultant bio-acoustic information as classification features. In an initial single-session study (N=25), we achieved mean Equal Error Rates (EERs) of 5.68% and 7.95% with microphones on the brow and left mastoid process. Combining the two signals substantially improves performance (to 2.35% EER). A subsequent multi-session study (N=30) demonstrates EERs are maintained over three recalls and, additionally, shows robustness to donning variations and background noise (achieving 2.72% EER). In a follow-up usability study over one week, participants report high levels of usability (as expressed by SUS scores) and that only modest workload is required to authenticate. Finally, a security analysis demonstrates the system’s robustness to spoofing and imitation attacks.

The gap between pre- and post-simulation, as well as the considerable layout time, increases the significance of the post-layout waveform prediction in dynamic random access memory (DRAM) design. This study develops a post-layout prediction model using the following two-stream neural network: (1) a multi-layer perceptron neural network to calculate the coupling noise by using the physical properties of global interconnects, and (2) a convolutional neural network to compute the time series trends of the waveforms by referencing adjacent signals. The proposed model trains two types of heterogeneous data such that accuracy of 95.5% is achieved on the 1b DRAM process 16Gb DDR5 composed of hundreds of millions of transistors. The model significantly improves the design completeness by pre-detecting the deterioration in the signal quality via post-layout waveform prediction. Generally, although a few weeks are required to obtain post-layout waveforms after the circuit design process, waveforms can be instantly predicted using our proposed model.

Twitter is a popular public source for threat hunting. Many security vendors and security professionals use Twitter in practice for collecting Indicators of Compromise (IOCs). However, little is known about IOCs on Twitter. Their important characteristics such as earliness, uniqueness, and accuracy have never been investigated. Moreover, how to extract IOCs from Twitter with high accuracy is not obvious. In this paper, we present Twiti, a system that automatically extracts various forms of malware IOCs from Twitter. Based on the collected IOCs, we conduct the first empirical assessment and thorough analysis of malware IOCs on Twitter. Twiti extracts IOCs from tweets identified as having malware IOC information by leveraging natural language processing and machine learning techniques. With extensive evaluation, we demonstrate that not only can Twiti extract malware IOCs accurately, but also the extracted IOCs are unique and early. By analyzing IOCs in Twiti from various aspects, we find that Twitter captures ongoing malware threats such as Emotet variants and malware distribution sites better than other public threat intelligence (TI) feeds. We also find that only a tiny fraction of IOCs on Twitter come from commercial vendor accounts and individual Twitter users are the main contributors of the early detected or exclusive IOCs, which indicates that Twitter can provide many valuable IOCs uncovered in commercial domain

There is plenty of threat-related information in open data sources. Early identification of emerging security threats from such information is an important part of security for deployed software and systems. While several cybersecurity event detection methods have been proposed to extract security events from unstructured text in open data sources, most of the existing methods focus on detecting events that have a large volume of mentions. On the contrary, to respond faster than attackers, security analysts and IT operators need to be aware of critical security events as early as possible, no matter how many mentions about an event are made. In this paper, we propose a novel event detection system that can quickly identify critical security events, such as new threats and resurgence of an attack or related event, from Twitter regardless of their volume of mentions. Unlike the existing methods, the proposed method triggers events by monitoring new words and re-emerging words, making it possible to narrow down candidate events among several hundreds of events. It then forms events by clustering tweets linked with the trigger words. This approach enables us to detect new and resurgent threats as early as possible. We empirically demonstrate that our system works promisingly over a wide range of threat types.

Keystroke data collected from smart devices includes various sensitive information about users. Collecting and analyzing such data raise serious privacy concerns. Google and Apple have recently applied local differential privacy (LDP) to address privacy issue on learning new words from users' keystroke data. However, these solutions require multiple LDP reports for a single word, which result in inefficient use of privacy budget and high computational cost. In this paper, we develop a novel algorithm for learning new words under LDP. Unlike the existing solutions, the proposed method generates only one LDP report for a single word. This enables the proposed method to use full privacy budget for generating a report and brings the benefit that the proposed method provides better utility at the same privacy degree than the existing methods. In our algorithm, each user appends a hash value to new word and sends only one LDP report of an n-gram selected randomly from the string packed by each new word and its hash value. The server then decodes frequent n-grams at each position of the string and discovers the candidate words by exploring graph-theoretic links between n-grams and checking integrity of candidates with hash values. Frequencies of frequent new words discovered are estimated from distribution estimates of n-grams by robust regression. We theoretically show that our algorithm can recover popular new words even though the server does not know the domain of the raw data. In addition, we theoretically and empirically demonstrate that our algorithm achieves higher accuracy compared to the existing solutions.

Local differential privacy (LDP) is a recently proposed privacy standard for collecting and analyzing data, which has been used, e.g., in the Chrome browser, iOS and macOS. In LDP, each user perturbs her information locally, and only sends the randomized version to an aggregator who performs analyses, which protects both the users and the aggregator against private information leaks. Although LDP has attracted much research attention in recent years, the majority of existing work focuses on applying LDP to complex data and/or analysis tasks. In this paper, we point out that the fundamental problem of collecting multidimensional data under LDP has not been addressed sufficiently, and there remains much room for improvement even for basic tasks such as computing the mean value over a single numeric attribute under LDP. Motivated by this, we first propose novel LDP mechanisms for collecting a numeric attribute, whose accuracy is at least no worse (and usually better) than existing solutions in terms of worst-case noise variance. Then, we extend these mechanisms to multidimensional data that can contain both numeric and categorical attributes, where our mechanisms always outperform existing solutions regarding worst-case noise variance. As a case study, we apply our solutions to build an LDP-compliant stochastic gradient descent algorithm (SGD), which powers many important machine learning tasks. Experiments using real datasets confirm the effectiveness of our methods, and their advantages over existing solutions.

A mobile operating system often needs to collect frequent new terms from users in order to build and maintain a comprehensive dictionary. Collecting keyboard usage data, however, raises privacy concerns. Local differential privacy (LDP) has been established as a strong privacy standard for collecting sensitive information from users. Currently, the best known solution for LDP-compliant frequent term discovery transforms the problem into collecting n-grams under LDP, and subsequently reconstructs terms from the collected n-grams by modelling the latter into a graph, and identifying cliques on this graph. Because the transformed problem (i.e., collecting n-grams) is very different from the original one (discovering frequent terms), the end result has poor utility. Further, this method is also rather expensive due to clique computation on a large graph. In this paper we tackle the problem head on: our proposal, PrivTrie, directly collects frequent terms from users by iteratively constructing a trie under LDP. While the methodology of building a trie is an obvious choice, obtaining an accurate trie under LDP is highly challenging. PrivTrie achieves this with a novel adaptive approach that conserves privacy budget by building internal nodes of the trie with the lowest level of accuracy necessary. Experiments using real datasets confirm that PrivTrie achieves high accuracy on common privacy levels, and consistently outperforms all previous methods.

Recommender systems are collecting and analyzing user data to provide better user experience. However, several privacy concerns have been raised when a recommender knows user's set of items or their ratings. A number of solutions have been suggested to improve privacy of legacy recommender systems, but the existing solutions in the literature can protect either items or ratings only. In this paper, we propose a recommender system that protects both user's items and ratings. For this, we develop novel matrix factorization algorithms under local differential privacy (LDP). In a recommender system with LDP, individual users randomize their data themselves to satisfy differential privacy and send the perturbed data to the recommender. Then, the recommender computes aggregates of the perturbed data. This framework ensures that both user's items and ratings remain private from the recommender. However, applying LDP to matrix factorization typically raises utility issues with i) high dimensionality due to a large number of items and ii) iterative estimation algorithms. To tackle these technical challenges, we adopt dimensionality reduction technique and a novel binary mechanism based on sampling. We additionally introduce a factor that stabilizes the perturbed gradients. With MovieLens and LibimSeTi datasets, we evaluate recommendation accuracy of our recommender system and demonstrate that our algorithm performs better than the existing differentially private gradient descent algorithm for matrix factorization under stronger privacy requirements.