Spam emails have been traditionally seen as just annoying and unsolicited emails containing advertisements, but they increasingly include scams, malware or phishing. In order to ensure the security and integrity for the users, organisations and researchers aim to develop robust filters for spam email detection. Recently, most spam filters based on machine learning algorithms published in academic journals report very high performance, but users are still reporting a rising number of frauds and attacks via spam emails. Two main challenges can be found in this field: (a) it is a very dynamic environment prone to the dataset shift problem and (b) it suffers from the presence of an adversarial figure, i.e. the spammer. Unlike classical spam email reviews, this one is particularly focused on the problems that this constantly changing environment poses. Moreover, we analyse the different spammer strategies used for contaminating the emails, and we review the state-of-the-art techniques to develop filters based on machine learning. Finally, we empirically evaluate and present the consequences of ignoring the matter of dataset shift in this practical field. Experimental results show that this shift may lead to severe degradation in the estimated generalisation performance, with error rates reaching values up to $$48.81\%$$48.81 % .

Face recognition is a valuable forensic tool for criminal investigators since it certainly helps in identifying individuals in scenarios of criminal activity like fugitives or child sexual abuse. It is, however, a very challenging task as it must be able to handle low-quality images of real world settings and fulfill real time requirements. Deep learning approaches for face detection have proven to be very successful but they require large computation power and processing time. In this work, we evaluate the speed–accuracy tradeoff of three popular deep-learning-based face detectors on the WIDER Face and UFDD data sets in several CPUs and GPUs. We also develop a regression model capable to estimate the performance, both in terms of processing time and accuracy. We expect this to become a very useful tool for the end user in forensic laboratories in order to estimate the performance for different face detection options. Experimental results showed that the best speed–accuracy tradeoff is achieved with images resized to $50\%$ of the original size in GPUs and images resized to $25\%$ of the original size in CPUs. Moreover, performance can be estimated using multiple linear regression models with a Mean Absolute Error (MAE) of 0.113, which is very promising for the forensic field.

Industrial control systems depend heavily on security and monitoring protocols. Several tools are available for this purpose, which scout vulnerabilities and take screenshots of various control panels for later analysis. However, they do not adequately classify images into specific control groups, which is crucial for security-based tasks performed by manual operators. To solve this problem, we propose a pipeline based on deep learning to classify snapshots of industrial control panels into three categories: internet technologies, operation technologies, and others. More specifically, we compare the use of transfer learning and fine-tuning in convolutional neural networks (CNNs) pre-trained on ImageNet to select the best CNN architecture for classifying the screenshots of industrial control systems. We propose the critical infrastructure dataset (CRINF-300), which is the first publicly available information technology (IT)/operational technology (OT) snapshot dataset, with 337 manually labeled images. We used the CRINF-300 to train and evaluate eighteen different pipelines, registering their performance under CPU and GPU environments. We found out that the Inception-ResNet-V2 and VGG16 architectures obtained the best results on transfer learning and fine-tuning, with F1-scores of $0.9832$ and $0.9373$, respectively. In systems where time is critical and the GPU is available, we recommend using the MobileNet-V1 architecture, with an average time of $0.03$ s to process an image and with an F1-score of $0.9758$.

Retrieving text embedded within images is a challenging task in real-world settings. Multiple problems such as low-resolution and the orientation of the text can hinder the extraction of information. These problems are common in environments such as Tor Darknet and Child Sexual Abuse images, where text extraction is crucial in the prevention of illegal activities. In this work, we evaluate eight text recognizers and, to increase the performance of text transcription, we combine these recognizers with rectification networks and super-resolution algorithms. We test our approach on four state-of-the-art and two custom datasets (TOICO-1K and Child Sexual Abuse (CSA)-text, based on text retrieved from Tor Darknet and Child Sexual Exploitation Material, respectively). We obtained a 0.3170 score of correctly recognized words in the TOICO-1K dataset when we combined Deep Convolutional Neural Networks (CNN) and rectification-based recognizers. For the CSA-text dataset, applying resolution enhancements achieved a final score of 0.6960. The highest performance increase was achieved on the ICDAR 2015 dataset, with an improvement of 4.83% when combining the MORAN recognizer and the Residual Dense resolution approach. We conclude that rectification outperforms super-resolution when applied separately, while their combination achieves the best average improvements in the chosen datasets.