Reversible Watermarking in Deep Convolutional Neural Networks for Integrity Authentication

Deep convolutional neural networks have made outstanding contributions in many fields such as computer vision in the past few years and many researchers published well-trained network for downloading. But recent studies have shown serious concerns about integrity due to model-reuse attacks and backdoor attacks. In order to protect these open-source networks, many algorithms have been proposed such as watermarking. However, these existing algorithms modify the contents of the network permanently and are not suitable for integrity authentication. In this paper, we propose a reversible watermarking algorithm for integrity authentication. Specifically, we present the reversible watermarking problem of deep convolutional neural networks and utilize the pruning theory of model compression technology to construct a host sequence used for embedding watermarking information by histogram shift. As shown in the experiments, the influence of embedding reversible watermarking on the classification performance is less than 0.5% and the parameters of the model can be fully recovered after extracting the watermarking. At the same time, the integrity of the model can be verified by applying the reversible watermarking: if the model is modified illegally, the authentication information generated by original model will be absolutely different from the extracted watermarking information.

Hyperspectral and LiDAR data classification based on linear self-attention

An efficient linear self-attention fusion model is proposed in this paper for the task of hyperspectral image (HSI) and LiDAR data joint classification. The proposed method is comprised of a feature extraction module, an attention module, and a fusion module. The attention module is a plug-and-play linear self-attention module that can be extensively used in any model. The proposed model has achieved the overall accuracy of 95.40% on the Houston dataset. The experimental results demonstrate the superiority of the proposed method over other state-of-the-art models.

Completeness-resolvable graphs

Given a connected graph $G=(V(G), E(G))$, the length of a shortest path from a vertex $u$ to a vertex $v$ is denoted by $d(u,v)$. For a proper subset $W$ of $V(G)$, let $m(W)$ be the maximum value of $d(u,v)$ as $u$ ranging over $W$ and $v$ ranging over $V(G)setminus W$. The proper subset $W={w_1,ldots,w_{|W|}}$ is a {em completeness-resolving set} of $G$ if $$ Psi_W: V(G)setminus W longrightarrow [m(W)]^{|W|},qquad ulongmapsto (d(w_1,u),ldots,d(w_{|W|},u)) $$ is a bijection, where $$ [m(W)]^{|W|}={(a_{(1)},ldots,a_{(|W|)})mid 1leq a_{(i)}leq m(W)text{ for each }i=1,ldots,|W|}. $$ A graph is {em completeness-resolvable} if it admits a completeness-resolving set. In this paper, we first construct the set of all completeness-resolvable graphs by using the edge coverings of some vertices in given bipartite graphs, and then establish posets on some subsets of this set by the spanning subgraph relationship. Based on each poset, we find the maximum graph and give the lower and upper bounds for the number of edges in a minimal graph. Furthermore, minimal graphs satisfying the lower or upper bound are characterized.

High-Fidelity Transfer of 2D Bismuth Oxyselenide and its Mechanical Properties

Two-dimensional (2D) bismuth oxyselenide (Bi2O2Se) with high electron mobility is advantageous in future high-performance and flexible electronic and optoelectronic devices. However, transfer of thin Bi2O2Se flakes is rather challenging, restricting measurements of its mechanical properties and application exploration in flexible devices. Here, we develop a reliable and effective polydimethylsiloxane (PDMS)-mediated method that allows transferring thin Bi2O2Se flakes from grown substrates onto target substrates like micro-electro-mechanical system substrates. The high fidelity of the transferred thin flakes stems from the high adhesive energy and flexibility of PDMS film. For the first time, the mechanical properties of 2D Bi2O2Se are experimentally acquired with nanoindentation method. We found that few-layer Bi2O2Se exhibits a large intrinsic stiffness of 18-23 GPa among 2D semiconductors, and a Young s modulus of 88.7 +- 14.4 GPa which is consistent with the theoretical values. Furthermore, few-layer Bi2O2Se can withstand a high radial strain of more than 3%, demonstrating excellent flexibility. The development of the reliable transfer method and documentation of mechanical properties of 2D Bi2O2Se jointly fill the gap between theoretical prediction and experimental verification of mechanical properties of this emerging material, and will promote flexible electronics and optoelectronics based on 2D Bi2O2Se.

Synthesis of ultrahigh-quality monolayer molybdenum disulfide through in-situ defect healing with thiol molecules

Monolayer transition metal dichalcogenides (TMDCs) are two-dimensional (2D) materials with many potential applications. Chemical vapour deposition (CVD) is a promising method to synthesize these materials. However, CVD-grown materials generally have poorer quality than mechanically exfoliated ones and contain more defects due to the difficulties in controlling precursors distribution and concentration during growth where solid precursors are used. Here, we propose to use thiol as a liquid precursor for CVD growth of high quality and uniform 2D MoS2. Atomic-resolved structure characterizations indicate that the concentration of sulfur vacancies in the MoS2 grown from thiol is the lowest among all reported CVD samples. Low temperature spectroscopic characterization further reveals the ultrahigh optical quality of the grown MoS2. Density functional theory simulations indicate that thiol molecules could interact with sulfur vacancies in MoS2 and repair these defects during the growth of MoS2, resulting in high quality MoS2. This work provides a facile and controllable method for the growth of high-quality 2D materials with ultralow sulfur vacancies and high optical quality, which will benefit their optoelectronic applications.

High Yield Growth and Doping of Black Phosphorus with Tunable Electronic Properties

Black phosphorus (BP) has recently attracted significant interest due to its unique electronic and optical properties. Doping is an effective strategy to tune a materials electronic structures, however, the direct and controllable growth of BP with a high yield and its doping remain a great challenge. Here we report an efficient short-distance transport (SDT) growth approach and achieve the controlled growth of high quality BP with the highest yield so far, where 98% of the red phosphorus is converted to BP. The doping of BP by As, Sb, Bi, Se and Te are also achieved by this SDT growth approach. Spectroscopic results show that doping systematically changes its electronic structures including band gap, work function, and energy band position. As a result, we have found that the air-stability of doped BP samples (Sb and Te-doped BP) improves compared with pristine BP, due to the downshift of the conduction band minimum with doping. This work develops a new method to grow BP and doped BP with tunable electronic structures and improved stability, and should extend the uses of these class of materials in various areas.

Gate-modulated conductance of few-layer WSe_2 field-effect transistors in the subgap regime: Schottky barrier transistor and subgap impurity states

Two key subjects stand out in the pursuit of semiconductor research: material quality and contact technology. The fledging field of atomically thin transition metal dichalcogenides (TMDCs) faces a number of challenges in both efforts. This work attempts to establish a connection between the two by examining the gate-dependent conductance of few-layer (1-5L) WSe2 field effect devices. Measurements and modeling of the subgap regime reveal Schottky barrier transistor behavior. We show that transmission through the contact barrier is dominated by thermionic field emission (TFE) at room temperature, despite the lack of intentional doping. The TFE process arises due to a large number of subgap impurity states, the presence of which also leads to high mobility edge carrier densities. The density of states of such impurity states is self-consistently determined to be approximately 1-2x10^13 /cm^2/eV in our devices. We demonstrate that substrate is unlikely to be a major source of the impurity states and suspect that lattice defects within the material itself are primarily responsible. Our experiments provide key information to advance the quality and understanding of TMDC materials and electrical devices.

Manipulating the magnetic anisotropy of cobalt doped titanium dioxide by carrier accumulation

Based on first-principles calculations, we predict that the magnetic anisotropy energy (MAE) of Co-doped TiO$_2$ sensitively depends on carrier accumulation. This magnetoelectric phenomenon provides a promising route to directly manipulate the magnetization direction of diluted magnetic semiconductor by external electric-fields. We calculate the band structures and reveal the origin of carrier-dependent MAE in k-space. In fact, the carrier accumulation shifts the Fermi energy and regulates the competing contributions to MAE. The first-principles calculations provide a straightforward way to design spintronics materials with electrically controllable spin direction.

Ab initio study of magnetic anisotropy in cobalt doped zinc oxide with electron-filling

Based on first-principles calculation, it has been predicted that the magnetic anisotropy energy (MAE) in Co-doped ZnO (Co:ZnO) depends on electron-filling. Results show that the charge neutral Co:ZnO presents a easy plane magnetic state. While modifying the total number of electrons, the easy axis rotates from in-plane to out-of-plane. The alternation of the MAE is considered to be the change of the ground state of Co ion, resulting from the relocating of electrons on Co d-orbitals with electron-filling.