Toxic comment classification models are often found biased toward identity terms which are terms characterizing a specific group of people such as "Muslim" and "black". Such bias is commonly reflected in false-positive predictions, i.e. non-toxic comments with identity terms. In this work, we propose a novel approach to tackle such bias in toxic comment classification, leveraging the notion of subjectivity level of a comment and the presence of identity terms. We hypothesize that when a comment is made about a group of people that is characterized by an identity term, the likelihood of that comment being toxic is associated with the subjectivity level of the comment, i.e. the extent to which the comment conveys personal feelings and opinions. Building upon the BERT model, we propose a new structure that is able to leverage these features, and thoroughly evaluate our model on 4 datasets of varying sizes and representing different social media platforms. The results show that our model can consistently outperform BERT and a SOTA model devised to address identity term bias in a different way, with a maximum improvement in F1 of 2.43% and 1.91% respectively.

Insider attacks are one of the most challenging cybersecurity issues for companies, businesses and critical infrastructures. Despite the implemented perimeter defences, the risk of this kind of attack is still very high. In fact, the detection of insider attacks is a very complicated security task and presents a serious challenge to the research community. In this paper, we aim to address this issue by using deep learning algorithms Autoencoder and Variational Autoencoder deep. We will especially investigate the usefulness of applying these algorithms to automatically defend against potential internal threats, without human intervention. The effectiveness of these two models is evaluated on the public dataset CERT dataset (CERT r4.2). This version of the CERT Insider Threat Test dataset includes both benign and malicious activities generated from 1000 simulated users. The comparison results with other models show that the Variational Autoencoder neural network provides the best overall performance with a greater detection accuracy and a reasonable false positive rate

The rapid increase in the use of IoT devices brings many benefits to the digital society, ranging from improved efficiency to higher productivity. However, the limited resources and the open nature of these devices make them vulnerable to various cyber threats. A single compromised device can have an impact on the whole network and lead to major security and physical damages. This paper explores the potential of using network profiling and machine learning to secure IoT against cyber-attacks. The proposed anomaly-based intrusion detection solution dynamically and actively profiles and monitors all networked devices for the detection of IoT device tampering attempts as well as suspicious network transactions. Any deviation from the defined profile is considered to be an attack and is subject to further analysis. Raw traffic is also passed on to the machine learning classifier for examination and identification of potential attacks. Performance assessment of the proposed methodology is conducted on the Cyber-Trust testbed using normal and malicious network traffic. The experimental results show that the proposed anomaly detection system delivers promising results with an overall accuracy of 98.35% and 0.98% of false-positive alarms.

Different performance metrics available are used to evaluate the Machine Learning Algorithms.

In this post, I will implement some of the most common loss functions for image segmentation in Keras/TensorFlow. I will only consider the case of two classes (i.e. Due to numerical stability, it is always better to use BinaryCrossentropy with from_logits True. You can see in the original code that TensorFlow sometimes tries to compute cross entropy from probabilities (when from_logits False). Due to numerical instabilities clip_by_value becomes then necessary.

Modern neural networks are powerful predictive models. However, when it comes to recognizing that they may be wrong about their predictions, they perform poorly. For example, for one of the most common activation functions, the ReLU and its variants, even a well-calibrated model can produce incorrect but high confidence predictions. In the related task of action recognition, most current classification methods are based on clip-level classifiers that densely sample a given video for non-overlapping, same-sized clips and aggregate the results using an aggregation function - typically averaging - to achieve video level predictions. While this approach has shown to be effective, it is sub-optimal in recognition accuracy and has a high computational overhead. To mitigate both these issues, we propose the confidence distillation framework to teach a representation of uncertainty of the teacher to the student sampler and divide the task of full video prediction between the student and the teacher models. We conduct extensive experiments on three action recognition datasets and demonstrate that our framework achieves significant improvements in action recognition accuracy (up to 20%) and computational efficiency (more than 40%).

Distantly supervised relation extraction (RE) automatically aligns unstructured text with relation instances in a knowledge base (KB). Due to the incompleteness of current KBs, sentences implying certain relations may be annotated as N/A instances, which causes the so-called false negative (FN) problem. Current RE methods usually overlook this problem, inducing improper biases in both training and testing procedures. To address this issue, we propose a two-stage approach. First, it finds out possible FN samples by heuristically leveraging the memory mechanism of deep neural networks. Then, it aligns those unlabeled data with the training data into a unified feature space by adversarial training to assign pseudo labels and further utilize the information contained in them. Experiments on two wildly-used benchmark datasets demonstrate the effectiveness of our approach.

The ability to identify applications based on the network data they generate could be a valuable tool for cyber defense. We report on a machine learning technique capable of using netflow-like features to predict the application that generated the traffic. In our experiments, we used ground-truth labels obtained from host-based sensors deployed in a large enterprise environment; we applied random forests and multilayer perceptrons to the tasks of browser vs. non-browser identification, browser fingerprinting, and process name prediction. For each of these tasks, we demonstrate how machine learning models can achieve high classification accuracy using only netflow-like features as the basis for classification.

As algorithmic risk assessment instruments (RAIs) are increasingly adopted to assist decision makers, their predictive performance and potential to promote inequity have come under scrutiny. However, while most studies examine these tools in isolation, researchers have come to recognize that assessing their impact requires understanding the behavior of their human interactants. In this paper, building off of several recent crowdsourcing works focused on criminal justice, we conduct a vignette study in which laypersons are tasked with predicting future re-arrests. Our key findings are as follows: (1) Participants often predict that an offender will be rearrested even when they deem the likelihood of re-arrest to be well below 50%; (2) Participants do not anchor on the RAI's predictions; (3) The time spent on the survey varies widely across participants and most cases are assessed in less than 10 seconds; (4) Judicial decisions, unlike participants' predictions, depend in part on factors that are orthogonal to the likelihood of re-arrest. These results highlight the influence of several crucial but often overlooked design decisions and concerns around generalizability when constructing crowdsourcing studies to analyze the impacts of RAIs.

Dry eye disease (DED) has a prevalence of between 5 and 50\%, depending on the diagnostic criteria used and population under study. However, it remains one of the most underdiagnosed and undertreated conditions in ophthalmology. Many tests used in the diagnosis of DED rely on an experienced observer for image interpretation, which may be considered subjective and result in variation in diagnosis. Since artificial intelligence (AI) systems are capable of advanced problem solving, use of such techniques could lead to more objective diagnosis. Although the term `AI' is commonly used, recent success in its applications to medicine is mainly due to advancements in the sub-field of machine learning, which has been used to automatically classify images and predict medical outcomes. Powerful machine learning techniques have been harnessed to understand nuances in patient data and medical images, aiming for consistent diagnosis and stratification of disease severity. This is the first literature review on the use of AI in DED. We provide a brief introduction to AI, report its current use in DED research and its potential for application in the clinic. Our review found that AI has been employed in a wide range of DED clinical tests and research applications, primarily for interpretation of interferometry, slit-lamp and meibography images. While initial results are promising, much work is still needed on model development, clinical testing and standardisation.