Biblio

We introduce AdaMix, an adaptive differentially private algorithm for training deep neural network classifiers using both private and public image data. While pre-training language models on large public datasets has enabled strong differential privacy (DP) guarantees with minor loss of accuracy, a similar practice yields punishing trade-offs in vision tasks. A few-shot or even zero-shot learning baseline that ignores private data can outperform fine-tuning on a large private dataset. AdaMix incorporates few-shot training, or cross-modal zero-shot learning, on public data prior to private fine-tuning, to improve the trade-off. AdaMix reduces the error increase from the non-private upper bound from the 167–311% of the baseline, on average across 6 datasets, to 68-92% depending on the desired privacy level selected by the user. AdaMix tackles the trade-off arising in visual classification, whereby the most privacy sensitive data, corresponding to isolated points in representation space, are also critical for high classification accuracy. In addition, AdaMix comes with strong theoretical privacy guarantees and convergence analysis.

The Industrial Internet of Things (IIoT) also referred as Cyber Physical Systems (CPS) as critical elements, expected to play a key role in Industry 4.0 and always been vulnerable to cyber-attacks and vulnerabilities. Terrorists use cyber vulnerability as weapons for mass destruction. The dark web's strong transparency and hard-to-track systems offer a safe haven for criminal activity. On the dark web (DW), there is a wide variety of illicit material that is posted regularly. For supervised training, large-scale web pages are used in traditional DW categorization. However, new study is being hampered by the impossibility of gathering sufficiently illicit DW material and the time spent manually tagging web pages. We suggest a system for accurately classifying criminal activity on the DW in this article. Rather than depending on the vast DW training package, we used authorized regulatory to various types of illicit activity for training Machine Learning (ML) classifiers and get appreciable categorization results. Espionage, Sabotage, Electrical power grid, Propaganda and Economic disruption are the cyber warfare motivations and We choose appropriate data from the open source links for supervised Learning and run a categorization experiment on the illicit material obtained from the actual DW. The results shows that in the experimental setting, using TF-IDF function extraction and a AdaBoost classifier, we were able to achieve an accuracy of 0.942. Our method enables the researchers and System authoritarian agency to verify if their DW corpus includes such illicit activity depending on the applicable rules of the illicit categories they are interested in, allowing them to identify and track possible illicit websites in real time. Because broad training set and expert-supplied seed keywords are not required, this categorization approach offers another option for defining illicit activities on the DW.

Neural architectures are the foundation for improving performance of deep neural networks (DNNs). This paper presents deep compositional grammatical architectures which harness the best of two worlds: grammar models and DNNs. The proposed architectures integrate compositionality and reconfigurability of the former and the capability of learning rich features of the latter in a principled way. We utilize AND-OR Grammar (AOG) as network generator in this paper and call the resulting networks AOGNets. An AOGNet consists of a number of stages each of which is composed of a number of AOG building blocks. An AOG building block splits its input feature map into N groups along feature channels and then treat it as a sentence of N words. It then jointly realizes a phrase structure grammar and a dependency grammar in bottom-up parsing the “sentence” for better feature exploration and reuse. It provides a unified framework for the best practices developed in state-of-the-art DNNs. In experiments, AOGNet is tested in the ImageNet-1K classification benchmark and the MS-COCO object detection and segmentation benchmark. In ImageNet-1K, AOGNet obtains better performance than ResNet and most of its variants, ResNeXt and its attention based variants such as SENet, DenseNet and DualPathNet. AOGNet also obtains the best model interpretability score using network dissection. AOGNet further shows better potential in adversarial defense. In MS-COCO, AOGNet obtains better performance than the ResNet and ResNeXt backbones in Mask R-CNN.

Adversarial attacks to image classification systems present challenges to convolutional networks and opportunities for understanding them. This study suggests that adversarial perturbations on images lead to noise in the features constructed by these networks. Motivated by this observation, we develop new network architectures that increase adversarial robustness by performing feature denoising. Specifically, our networks contain blocks that denoise the features using non-local means or other filters; the entire networks are trained end-to-end. When combined with adversarial training, our feature denoising networks substantially improve the state-of-the-art in adversarial robustness in both white-box and black-box attack settings. On ImageNet, under 10-iteration PGD white-box attacks where prior art has 27.9% accuracy, our method achieves 55.7%; even under extreme 2000-iteration PGD white-box attacks, our method secures 42.6% accuracy. Our method was ranked first in Competition on Adversarial Attacks and Defenses (CAAD) 2018 — it achieved 50.6% classification accuracy on a secret, ImageNet-like test dataset against 48 unknown attackers, surpassing the runner-up approach by 10%. Code is available at https://github.com/facebookresearch/ImageNet-Adversarial-Training.

During the last years, criminals have become aware of how digital evidences that lead them to courts and jail are collected and analyzed. Hence, they have started to develop antiforensic techniques to evade, hamper, or nullify their evidences. Nowadays, these techniques are broadly used by criminals, causing the forensic analysis to be in a state of decay. To defeat against these techniques, forensic analyst need to first identify them, and then to mitigate somehow their effects. In this paper, wereview the anti-forensic techniques and propose a new taxonomy that relates them to the initial phase of a forensic process mainly affected by each technique. Furthermore, we introduce mitigation techniques for these anti-forensic techniques, considering the chance to overcome the anti-forensic techniques and the difficulty to apply them.