Important concerns when using medical data for Machine Learning (ML) is patient privacy and bias. Federated Learning (FL), the training of a centralized model by using parameters from decentralized models, is alternatively used to protect patient privacy. Medical data can often be structured and sparse, where deep learning is not applicable. In this work, we applied Federated Learning with Logistic Regression on medical data through multiple experiments with data distribution among clients. Three simulated cluster sampling methods conducted to compare model accuracy with different data distributions and/or sample sizes. Our observations were as follows: (1) the Federated Learning model performs, on average, better than the average of individual clients, and (2) variability increases as the sample size and the number of clients increases, and accuracy stays moderately consistent. In addition, we designed a Federated Learning system that securely transfers model parameters between server and clients. We also present an approach to study bias in these models. Our results show that Federated Learning is a promising approach for medical applications.

According to the International Labor Organization (ILS), approximately 27.6 million people were victims of human exploitation in the year 2021. Trafficker adaptability, victim vulnerability, and technological challenges make it difficult to capture the perpetrators of these heinous crimes. The crime typically occurs in hotel rooms, where traffickers capture images of their victims for use in illegal advertisements and other criminal activities. However, there are several challenges to identifying the hotels from these images, such as poor quality of hotel images and assorted camera angles. Different hotel chains with similar room types can have homogeneity in room designs, leading to low inter-class variation. The existing solutions lack the complexity and efficiency to correctly classify the hotel rooms. Our goal is to develop a comprehensive system that enables law enforcement authorities to curb human trafficking. The proposed solution enables the user to find the top-5 hotels that are most similar to the one in the image drawn from the dark web or ongoing cases. The system aids in tracking, verification, and addition of case-related information to the database. It serves as a human-in-the-loop system by providing a wide set of features like crime stat reports, visualizations, and map views. The end goal of the system is to expedite and streamline the investigation process for human trafficking by catching the perpetrators involved in these heinous crimes and delivering freedom to the victims.

Depth estimation algorithms are useful components of computer vision systems to assess video streams on urban streets. They can provide important information about the street space and improve situational awareness for humans using the street space. Deep learning algorithms for depth estimation algorithms are slow for these applications. To provide situational awareness to humans, the inference time needs to be small, so that results are fresh and meaningful. This paper explores two promising approaches - pruning and quantization - to improve inference time with a little compromise on the performance. We explore the impact of each of these methods independently and introduce a hybrid method that performs pruning followed by quantization. We evaluate the execution time, resource utilization, and performance of various state-of-the-art depth estimation algorithms with and without one or both of these techniques. We observe that using both pruning and quantization can improve the inference time dramatically, with a 39.6% speedup in inference time and an 81.2% reduction in memory utilization while losing only 4% in performance.

Cloud computing offers an economical and elastic means to handle the storage and computation needs of the Internet of Things (IoT). However, storage and retrieval from the cloud could potentially violate policies, especially those pertaining to data privacy. Edge computing as a paradigm is a suitable way to overcome these issues. This paper presents a policy-aware edge computing architecture that enables policy-aware normalization and filtration of the data that is sent to cloud services to preserve policies. We use a secure and encrypted channel to transmit the data generated by the IoT devices to the dedicated computing units at the edge of the network. Our architecture offers programmers the ability to configure the system easily and perform a predetermined set of computation tasks on the data, e.g., tasks to uphold privacy policies such as blurring faces, license plates, etc.