To adapt to the rapidly increasing vulnerabilities in software products and cyber threats that exploit them, security professionals are actively working with software developers to produce more secure systems. In software development, agile methods are increasingly adopted in critical software projects where security risks are prominent challenges. This adoption stems from the fact that agile methods are highly iterative and support delivering services and products in smaller batches which allows security professionals to seamlessly integrate software development security activities with agile methodologies. In addition, the iterative nature of agile software development encourages frequent inspections, tests, and patching of software systems to mitigate cybersecurity risks and vulnerabilities. Considering the massive growth of the Internet of Things (IoT) and Intelligent Transportation Systems (ITS) products, the challenge of software development while addressing the security and safety concerns of these devices will continue to increase. This paper presents a comprehensive and detailed review of agile software development in the context of IoT, ITS, and their cybersecurity and risk challenges. Furthermore, we provide a systematic comparison of the reviewed literature based on a set of defined criteria. Finally, we provide a broader outlook and an outline for designing future security-enhanced agile software development solutions for IoT and ITS systems.

Lung cancer is a common type of cancer that causes death if not detected early enough. Doctors use computed tomography (CT) images to diagnose lung cancer. The accuracy of the diagnosis relies highly on the doctor's expertise. Recently, clinical decision support systems based on deep learning valuable recommendations to doctors in their diagnoses. In this paper, we present several deep learning models to detect non-small cell lung cancer in CT images and differentiate its main subtypes namely adenocarcinoma, large cell carcinoma, and squamous cell carcinoma. We adopted standard convolutional neural networks (CNN), visual geometry group-16 (VGG16), and VGG19. Besides, we introduce a variant of the CNN that is augmented with convolutional block attention modules (CBAM). CBAM aims to extract informative features by combining cross-channel and spatial information. We also propose variants of VGG16 and VGG19 that utilize a support vector machine (SVM) at the classification layer instead of SoftMax. We validated all models in this study through extensive experiments on a CT lung cancer dataset. Experimental results show that supplementing CNN with CBAM leads to consistent improvements over vanilla CNN. Results also show that the VGG variants that use the SVM classifier outperform the original VGGs by a significant margin.

Social media sites are considered one of the most important sources of data in many fields, such as health, education, and politics. While surveys provide explicit answers to specific questions, posts in social media have the same answers implicitly occurring in the text. This research aims to develop a method for extracting implicit answers from large tweet collections, and to demonstrate this method for an important concern: the problem of heart attacks. The approach is to collect tweets containing “heart attack” and then select from those the ones with useful information. Informational tweets are those which express real heart attack issues, e.g., “Yesterday morning, my grandfather had a heart attack while he was walking around the garden.” On the other hand, there are non-informational tweets such as “Dropped my iPhone for the first time and almost had a heart attack.” The starting point was to manually classify around 7000 tweets as either informational (11%) or non-informational (89%), thus yielding a labeled dataset to use in devising a machine learning classifier that can be applied to our large collection of over 20 million tweets. Tweets were cleaned and converted to a vector representation, suitable to be fed into different machine-learning algorithms: Deep neural networks, support vector machine (SVM), J48 decision tree and naïve Bayes. Our experimentation aimed to find the best algorithm to use to build a high-quality classifier. This involved splitting the labeled dataset, with 2/3 used to train the classifier and 1/3 used for evaluation besides cross-validation methods. The deep neural network (DNN) classifier obtained the highest accuracy (95.2%). In addition, it obtained the highest F1-scores with (73.6%) and (97.4%) for informational and non-informational classes, respectively.