DeFi, a decentralized financial service based on blockchain, not only provides innovative financial services, but also poses various risks, such as the Terra Luna crash. Therefore, anomaly detection in DeFi is necessary to ensure the safety and reliability of the DeFi ecosystem. However, this is very difficult because of the complex protocol, interaction among smart contracts, and high market volatility. In this study, we propose a novel method to effectively detect anomalies in DeFi. To the best of our knowledge, this is the first study that utilizes deep learning to detect anomalies in DeFi. We propose a deep learning model, anomaly VAE-Transformer, which combines the variational autoencoder to extract local information in the short term, and the transformer, to identify dependencies between data in the long term. Based on a deep understanding of DeFi protocols, the proposed model collects and analyzes various on-chain data of Olympus DAO, a representative DeFi protocol, for extracting features suitable for anomaly detection. Then, we demonstrate the superiority of the proposed model by analyzing four anomaly cases detected successfully by the proposed model in Olympus DAO. A malicious attack attempt and structural changes in DeFi protocols can be identified quickly using the proposed method; this is expected to help protect the assets of DeFi users and improve the safety, reliability, and transparency of the DeFi market. The dataset and codes are available at https://github.com/fialle/Anomaly-VAE-Transformer.

As violent criminals, such as child sex offenders, tend to have high recidivism rates in modern society, there is a need to prevent such offenders from approaching socially disadvantaged and crime-prone areas, such as schools or childcare centers. Accordingly, national governments and related institutions have installed surveillance cameras and provided additional personnel to manage and monitor them via video surveillance equipment. However, naked-eye monitoring by guards and manual image processing cannot properly evaluate the video captured by surveillance cameras. To address the various problems of conventional systems that simply store and retrieve image data, a system is needed that can actively classify captured images in real-time, in addition to assisting surveillance personnel. Therefore, this paper proposes a video surveillance system based on a composable deep face recognition method. The proposed system detects the faces of criminals in real time from videos captured by a surveillance camera and notifies relevant institutions of the appearance of criminals. For real-time face detection, a down-sampled image forked from the original is used to localize unspecified faces. To improve accuracy and confidence in the recognition task, a scoring method based on face tracking is proposed. The final score combines the recognition confidence and the standard score to determine the embedding distance from the criminal face embedding data. The blind spots of surveillance personnel can be effectively addressed through early detection of criminals approaching crime-prone areas. The contributions of the paper are as follows. The proposed system can process images from surveillance cameras in real-time by using down-sampling. It can effectively identify the identity of criminals by using a face tracking ID unit and minimizes prediction reversal by solving the congested embedding problem in the feature space that may occur when performing identification matching on a large amount of face embedding DBs. Additionally, the reliability of the identification results is complemented by an identification score accumulation method. In this paper, we prototyped the proposed system and experimented with the recognition model, achieving an accuracy of 0.900 and an F-1 score of 0.943. We also experimentally confirmed that the models proposed in other studies have higher performance when using the tracked instance-level face identification method proposed in this paper. It is expected that the proposed system can be used to locate criminals and protect national facilities, and such responses can quickly prevent accidents/incidents. The dataset and code are available at https://github.com/aengoo/focusface.

Blockchain and artificial intelligence are two of the most prominent technologies in computer science today and have attracted considerable attention from various research communities. Recently, several initiatives have been launched to explore the combination of these two pioneering technologies. The main goal is to combine the data integrity, privacy, and decentralization properties of blockchain with the ability of artificial intelligence to process, analyze, predict, and refine massive data sets. The combination of blockchain and AI technologies is expected to address key challenges in the digital realm, such as data security, transparency, and streamlined decision-making. However, there is a problem that many studies have focused on the advancement of a single technology as the main perspective. To overcome these recent research limitations, we provide a broad view of the combination of blockchain and artificial intelligence and analyze the limitations of existing research and their causes. Furthermore, we identify challenges and attempts to be addressed through this analysis. The analysis in this paper is organized into a comprehensive section dedicated to the application of artificial intelligence in blockchain and vice versa. Based on our analysis, we identify existing challenges and propose a novel framework for researchers to overcome these limitations, thus expanding new research opportunities.