As deepfake becomes more sophisticated, the demand for fake facial image detection is increasing. Although great progress has been made in detection, performance of most existing methods degrade significantly when these are applied to detect low-quality images disappearance key clues during compression process. In this work, we mine frequency domain and RGB information specifically improve compressed images. Our method consists two modules: (1) a preprocessing module (2) classification...

Rust aims to offer full memory safety for programs, a guarantee that untamed C programs do not enjoy. How difficult is it translate existing code Rust? To get complementary view from of automatic translators, we report on user study asking humans real-world Rust. Our participants are able produce safe translations, whereas state-of-the-art tools so. analysis highlights the high-level strategy taken by users departs significantly those study. We also find often choose zero-cost (static)...

Face tampering is an intriguing task in video/image genuineness identification and has attracted significant amounts of attention recent years. In this work, we propose a face forgery detection method that consists preprocessing, improved Siamese network-based feature extractor (including alignment module), postprocessing (a voting principle). Roughly speaking, our extracts the features grey space face/background image pairs measures difference to make decisions. Experiments on several...

In recent years, there has been increasing interest in studying the adversarial attack, which poses potential risks to deep learning applications and stimulated numerous researches, e.g. improving robustness of neural networks. this work, we propose a novel double-stream architecture – Guided Erasable Adversarial Attack (GEAA) for protecting high-quality labeled data with high commercial values under data-sharing scenarios. GEAA contains three phases, denoising reconstruction,...

Recent studies highlight the vulnerability of convolutional neural networks (CNNs) to adversarial attacks, which also calls into question reliability forensic methods. Existing attacks generate one-to-one noise, means these methods have not learned fingerprint information. Therefore, we introduce two powerful copy-move attack, and joint feature-based auto-learning attack. To validate performance attack methods, move a step ahead higher possible defense mechanism relation mismatch. expands...

Cell-phones have become a necessary communication accessory in daily life. MMS (Multimedia Messaging Service) used by smart phones has caused higher requirement on mobile image manipulation. Classifying source cell-phones major issue the cell-phone forensics. There are two ways usually for tracing and identifying device: characteristics equipment fingerprint. Both of above schemes require set images captured known training classification model. To avoid using any prior knowledge practical...

Double compression usually occurs after the image has gone through some kinds of tampering, so double detection is a basic mean to assess authenticity given image. In this paper, we propose model distribution mode based first digits DCT (Discrete Cosine Transform) coefficients using Markov transition probability matrix and utilize its stationary as features for detection. Experiment results show effectiveness proposed method comparison been made improvement by second order statistical model.

Traditional optimization methods of heated oil pipeline operation are usually offline or nondynamic, where steady-state models complex and inaccurate physical utilized. To achieve online with high adaptability timeliness, this paper deploys a neural network (NN) to identify imitate real thermal pressure response system instead solving complicated model equations. The NN is trained built by history data obtained from the Supervisory Control Data Acquisition (SCADA) system. For certain...

A modified Fourier–Ritz approach is developed in this study to analyze the free in-plane vibration of orthotropic annular sector plates with general boundary conditions. In approach, two auxiliary sine functions are added standard Fourier cosine series obtain a robust function set. The introduction logarithmic radial variable simplifies expressions total energy and Lagrangian function. improved expansion based on new eliminates all potential discontinuities original displacement its...

Deep neural networks (DNNs) have seen extensive studies on image recognition and classification, segmentation, related topics. However, recent show that DNNs are vulnerable in defending adversarial examples. The classification network can be deceived by adding a small amount of perturbation to clean samples. There challenges when researchers want design general approach defend against wide variety To solve this problem, we introduce defensive method prevent examples from generating. Instead...

Nowadays, source camera identification, which aims to identify the of images, is quite important in field forensics. There a problem that cannot be ignored existing methods are unreliable and even out work case small training sample. To solve this problem, virtual sample generation-based method proposed paper, combined with ensemble learning. In after constructing sub-sets LBP features, authors generate sample-based on mega-trend-diffusion (MTD) method, calculates diffusion range samples...

Source camera identification is an important branch in the field of digital forensics. Most existing works are based on assumption that number training samples sufficient. However, practice, it unrealistic to obtain a large amount labeled samples. Therefore, order solve problem low accuracy for methods few-shot scenario, we propose novel method called prototype construction with ensemble projection (PCEP). In this work, extract variety features from datasets rich prior information. Then,...

Source Camera Identification (SCI) has been playing an important role in the security field for decades. With development of Deep Learning, performance SCI noteworthily improved. However, most proposed methods are forensic only a single camera identification category, e.g., model identification. For exploiting coupling between different categories, we present new coding method. That is, apply multi-task training method to regress namely, classify brands, models and devices synchronously...