Clustering Approach for Detecting Multiple Types of Adversarial Examples.

With intentional feature perturbations to a deep learning model, the adversary generates an adversarial example to deceive the deep learning model. As an adversarial example has recently been considered in the most severe problem of deep learning technology, its defense methods have been actively studied. Such effective defense methods against adversarial examples are categorized into one of the three architectures: (1) model retraining architecture; (2) input transformation architecture; and (3) adversarial example detection architecture. Especially, defense methods using adversarial example detection architecture have been actively studied. This is because defense methods using adversarial example detection architecture do not make wrong decisions for the legitimate input data while others do. In this paper, we note that current defense methods using adversarial example detection architecture can classify the input data into only either a legitimate one or an adversarial one. That is, the current defense methods using adversarial example detection architecture can only detect the adversarial examples and cannot classify the input data into multiple classes of data, i.e., legitimate input data and various types of adversarial examples. To classify the input data into multiple classes of data while increasing the accuracy of the clustering model, we propose an advanced defense method using adversarial example detection architecture, which extracts the key features from the input data and feeds the extracted features into a clustering model. From the experimental results under various application datasets, we show that the proposed method can detect the adversarial examples while classifying the types of adversarial examples. We also show that the accuracy of the proposed method outperforms the accuracy of recent defense methods using adversarial example detection architecture.

Is Homomorphic Encryption-Based Deep Learning Secure Enough?

As the amount of data collected and analyzed by machine learning technology increases, data that can identify individuals is also being collected in large quantities. In particular, as deep learning technology-which requires a large amount of analysis data-is activated in various service fields, the possibility of exposing sensitive information of users increases, and the user privacy problem is growing more than ever. As a solution to this user's data privacy problem, homomorphic encryption technology, which is an encryption technology that supports arithmetic operations using encrypted data, has been applied to various field including finance and health care in recent years. If so, is it possible to use the deep learning service while preserving the data privacy of users by using the data to which homomorphic encryption is applied? In this paper, we propose three attack methods to infringe user's data privacy by exploiting possible security vulnerabilities in the process of using homomorphic encryption-based deep learning services for the first time. To specify and verify the feasibility of exploiting possible security vulnerabilities, we propose three attacks: (1) an adversarial attack exploiting communication link between client and trusted party; (2) a reconstruction attack using the paired input and output data; and (3) a membership inference attack by malicious insider. In addition, we describe real-world exploit scenarios for financial and medical services. From the experimental evaluation results, we show that the adversarial example and reconstruction attacks are a practical threat to homomorphic encryption-based deep learning models. The adversarial attack decreased average classification accuracy from 0.927 to 0.043, and the reconstruction attack showed average reclassification accuracy of 0.888, respectively.

Robust Indoor Localization Methods Using Random Forest-Based Filter against MAC Spoofing Attack.

With the development of wireless networks and mobile devices, interest on indoor localization systems (ILSs) has increased. In particular, Wi-Fi-based ILSs are widely used because of the good prediction accuracy without additional hardware. However, as the prediction accuracy decreases in environments with natural noise, some studies were conducted to remove it. So far, two representative methods, i.e., the filtering-based method and deep learning-based method, have shown a significant effect in removing natural noise. However, the prediction accuracy of these methods severely decreased under artificial noise caused by adversaries. In this paper, we introduce a new media access control (MAC) spoofing attack scenario injecting artificial noise, where the prediction accuracy of Wi-Fi-based indoor localization system significantly decreases. We also propose a new deep learning-based indoor localization method using random forest(RF)-filter to provide the good prediction accuracy under the new MAC spoofing attack scenario. From the experimental results, we show that the proposed indoor localization method provides much higher prediction accuracy than the previous methods in environments with artificial noise.

Switching Magnetism and Superconductivity with Spin-Polarized Current in Iron-Based Superconductor.

We explore a new mechanism for switching magnetism and superconductivity in a magnetically frustrated iron-based superconductor using spin-polarized scanning tunneling microscopy (SPSTM). Our SPSTM study on single-crystal Sr_{2}VO_{3}FeAs shows that a spin-polarized tunneling current can switch the Fe-layer magnetism into a nontrivial C_{4} (2×2) order, which cannot be achieved by thermal excitation with an unpolarized current. Our tunneling spectroscopy study shows that the induced C_{4} (2×2) order has characteristics of plaquette antiferromagnetic order in the Fe layer and strongly suppresses superconductivity. Also, thermal agitation beyond the bulk Fe spin ordering temperature erases the C_{4} state. These results suggest a new possibility of switching local superconductivity by changing the symmetry of magnetic order with spin-polarized and unpolarized tunneling currents in iron-based superconductors.

Versatile variable temperature and magnetic field scanning probe microscope for advanced material research.

We have built a variable temperature scanning probe microscope (SPM) that covers 4.6 K-180 K and up to 7 T whose SPM head fits in a 52 mm bore magnet. It features a temperature-controlled sample stage thermally well isolated from the SPM body in good thermal contact with the liquid helium bath. It has a 7-sample-holder storage carousel at liquid helium temperature for systematic studies using multiple samples and field emission targets intended for spin-polarized spectroscopic-imaging scanning tunneling microscopy (STM) study on samples with various compositions and doping conditions. The system is equipped with a UHV sample preparation chamber and mounted on a two-stage vibration isolation system made of a heavy concrete block and a granite table on pneumatic vibration isolators. A quartz resonator (qPlus)-based non-contact atomic force microscope (AFM) sensor is used for simultaneous STM/AFM operation for research on samples with highly insulating properties such as strongly underdoped cuprates and strongly correlated electron systems.

Correlation of Fe-Based Superconductivity and Electron-Phonon Coupling in an FeAs/Oxide Heterostructure.

Interfacial phonons between iron-based superconductors (FeSCs) and perovskite substrates have received considerable attention due to the possibility of enhancing preexisting superconductivity. Using scanning tunneling spectroscopy, we studied the correlation between superconductivity and e-ph interaction with interfacial phonons in an iron-based superconductor Sr_{2}VO_{3}FeAs (T_{c}≈33 K) made of alternating FeSC and oxide layers. The quasiparticle interference measurement over regions with systematically different average superconducting gaps due to the e-ph coupling locally modulated by O vacancies in the VO_{2} layer, and supporting self-consistent momentum-dependent Eliashberg calculations provide a unique real-space evidence of the forward-scattering interfacial phonon contribution to the total superconducting pairing.

Malignant fibrous histiocytoma of the maxillary sinus in a spray painter from an automobile repair shop.

BACKGROUND: We report a case of a spray painter who developed malignant fibrous histiocytoma (MFH) of the maxillary sinus following long-term exposure to chromium, nickel, and formaldehyde, implying that these agents are probable causal agents of MFH. CASE REPORT: The patient developed right-sided prosopalgia that began twenty months ago. The symptom persisted despite medical treatment. After two months, he was diagnosed with MFH through imaging studies, surgery, and pathological microscopic findings at a university hospital in Seoul. His social, medical, and family history was unremarkable.The patient had worked for about 18 years at an automobile repair shop as a spray painter. During this period, he had been exposed to various occupational agents, such as hexavalent chromium, nickel, and formaldehyde, without appropriate personal protective equipment. He painted 6 days a week and worked for about 8 hours a day.Investigation of the patient's work environment detected hexavalent chromium, chromate, nickel, and formaldehyde. CONCLUSIONS: The study revealed that the patient had been exposed to hexavalent chromium, formaldehyde, and nickel compounds through sanding and spray painting. The association between paranasal cancer and exposure to the aforementioned occupational human carcinogens has been established. We suggest, in this case, the possibility that the paint spraying acted as a causal agent for paranasal cancer.