ARM (United Kingdom)

Smart city applications require machine learning (ML) capabilities at the edge, for which they require edge devices equipped with specialized processors (e.g., GPUs). However, deploying such high-end edge devices in large numbers is challenging, particularly in regions that face infrastructure and budget constraints (e.g., cities lacking ubiquitous broadband) or extreme weather conditions (e.g., hot temperatures over sustained periods of time). Hence, custom ML frameworks are necessary that can support efficient and safe consolidation of multiple applications with diverse workloads on fewer edge servers. We present IceEdge, an inference serving framework for CPU-GPU based edge computing servers, which isolates colocated applications using existing container mechanisms and provides a deadline- and thermal-aware scheduler for the ML workloads. We implement IceEdge for a Dockerized environment on NVIDIA Jetson Xavier AGX platform. IceEdge outperforms state-of-the-art real-time inference serving frameworks Triton and DeepRT, improving the frame drop rate by up to 15% and reducing the temperature rise by around 12°C compared to Triton inference serving framework.

Embedded systems do not benefit from strong memory protection, because they are designed to minimize cost. At the same time, there is increasing pressure to connect embedded devices to the internet, where their vulnerable nature makes them routinely subject to compromise. This fundamental tension leads to the current status-quo where exploitable devices put individuals and critical infrastructure at risk.

Edge AI encompasses the technologies at the intersection of machine learning (ML) and embedded systems, enabling the development of intelligent applications on devices with minimal power consumption. These devices, although constrained by limited memory and computational resources, utilize sensors to interact with the physical environment and make decisions using ML algorithms. Edge AI thrives on the synergy between ML architectures and deployment platforms, as the design of one profoundly influences the efficiency and functionality of the other. This keynote paper delves into the rationale for employing microcontrollers and neural processing units (NPUs) designed for constrained use-cases as target platforms for Edge AI, explores the necessity of running ML locally, and highlights the opportunities and challenges these systems pose. It describes the main applications of Edge AI and provides references to a practical use case.

ABSTRACTBackgroundHepatitis B is one of the major global health issues, which presents a particularly severe challenge within the confines of African prisons, characterized by high rates of transmission and limited access to adequate healthcare. The prevalence of Hepatitis B in these settings represents a silent crisis.ObjectiveThis research highlights the critical public health emergency posed by Hepatitis B in African prisons, underscoring the need for urgent intervention and comprehensive strategies.MethodsA comprehensive literature search was conducted to investigate the public health challenges posed by Hepatitis B in African prisons. The search focused on peer‐reviewed articles, policy documents, and original literature published from 2000 to 2024. Databases including PubMed, Scopus, Web of Science, and Google Scholar were utilized.ResultsThe prison environment, marked by overcrowding, inadequate sanitation, and high‐risk behaviors, fosters the rapid spread of Hepatitis B. The transmission is further exacerbated by limited access to vaccination, insufficient screening programs, and a lack of awareness among inmates and prison staff. Consequently, the incidence of Hepatitis B in African prisons is significantly higher than in the general population, creating a reservoir of infection that poses a broader public health threat upon prisoners' release. Addressing this crisis requires a multifaceted approach.ConclusionThis research calls for immediate and sustained action to mitigate the Hepatitis B crisis in African prisons. By prioritizing this issue within public health agendas, we can reduce transmission rates, improve health outcomes for inmates, and protect broader community health. The urgency of addressing Hepatitis B in African prisons cannot be overstated, as it represents a critical juncture in the fight against infectious diseases in marginalized populations.

Purpose: Constraint-induced movement therapy (CIMT) is an evidence-based intervention for arm recovery after acquired brain injury. Clinician knowledge, time and confidence in delivering CIMT are established barriers to the routine use of CIMT in practice. CIMT delivery via telehealth is one option to help overcome these barriers. This study aimed to understand clinician experiences of using an educational website and if the education and online resources contributed to their self-reported use of constraint-induced movement therapy via telehealth (TeleCIMT) in practice. Materials and Methods: Data were collected from a purposive sample of therapists registered to use the TeleCIMT website and website analytics. An online survey explored participants' experience with CIMT delivery (both face to face and via telehealth), their perceptions of the website, and barriers and enablers to TeleCIMT implementation using the Capability, Opportunity, Motivation-Behaviour model. Website analytics were used to evaluate website traffic and resource use. Data were analysed using descriptive statistics (quantitative data) and content analysis (qualitative data). Results: Forty therapists responded to the survey; 72.5% (n = 29) of the respondents were occupational therapists, and 37.5% (n = 15) had delivered TeleCIMT. Most of the participants agreed that the website was easy to navigate (n = 26, 90%) and felt that they had the knowledge (n = 28, 96.6%) and skills (n = 24, 82.7%) to deliver TeleCIMT. The enablers to TeleCIMT included motivation to implement learnings from the website, confidence in delivering the programme, and the convenience of remote delivery. The perceived barriers to TeleCIMT use included limited access to technology and the availability of a client supporter to enable engagement in TeleCIMT. The resources used most frequently by the respondents were the participant preparation pack and participant programme pack. Shorter video learning modules (<11 min in duration) had greater engagement than longer video learning modules. Conclusions: Whilst online education and resources may enhance clinician knowledge of constraint-induced movement therapy and telehealth delivery, other barriers such as lack of technology access, may need to be addressed through additional learning and implementation strategies to support the routine use of TeleCIMT in practice.

Abstract ESBMC v7.7 improves the verification of concurrent C programs by incorporating techniques such as dynamic thread scheduling, incremental SMT solving, and partial order reduction (POR). These improvements enhance the tool’s performance, particularly in exploring complex multi-threaded executions. The new scheduler prioritizes higher-thread identifiers during context switches, which helps explore deeper program states. The use of incremental SMT solving and a refined POR algorithm reduces the exploration of unreachable interleavings and redundant states. These updates enable ESBMC to detect bugs faster, making it a more effective tool for ensuring the safety of multi-threaded applications.

Convolutional Neural Networks (CNNs) have been utilised in many image and video processing applications. The convolution operator, also known as a spatial filter, is usually a linear operation, but this linearity compromises essential features and details inherent in the non-linearity present in many applications. However, due to its slow processing, the use of a nonlinear spatial filter is a significant bottleneck in many software applications. Further, due to their complexity, they are difficult to accelerate in FPGA or VLSI architectures. This paper presents novel FPGA implementations of linear and nonlinear spatial filters. More specifically, the arithmetic computations are carried out in custom floating-point, enabling a tradeoff of precision and hardware compactness, reducing algorithm development time. Further, we show that it is possible to process video at a resolution of 1080p with a frame rate of 60 frames per second, using a low-cost FPGA board. Finally, we show that using a domain-specific language will allow the rapid prototyping of image processing algorithms in custom floating-point arithmetic, allowing non-experts to quickly develop real-time video processing applications. INDEX TERMS Domain-specific language, embedded video processing, floating-point arithmetic, FPGA, real-time.

Serverless computing is a popular software deployment model for the cloud, in which applications are designed as a collection of stateless tasks. Developers are charged for the CPU time and memory footprint during the execution of each serverless function, which incentivizes them to reduce both runtime and memory usage. As a result, functions tend to be short (often on the order of a few milliseconds) and compact (128–256 MB). Cloud providers can pack thousands of such functions on a server, resulting in frequent context switches and a tremendous degree of interleaving. As a result, when a given memory-resident function is re-invoked, it commonly finds its on-chip microarchitectural state completely cold due to thrashing by other functions — a phenomenon termed lukewarm invocation.

Ucam is a user-centric, blockchain-based and end-to-end secure home IP camera system designed by IoTeX. In a Ucam system, all the video clips captured by the camera are encrypted using a user-controlled symmetric key before storing them in the cloud. In this demo paper, we describe the system architecture and implementation of a privacy-preserving object detection proof-of-concept (PoC) for the Ucam system using Veracruz, a confidential computing framework developed by Arm Research. The resulting PoC demonstrates the viability of bringing the object detection capability to the Ucam system without leaking users sensitive information.

Rowhammer is a serious security problem of contemporary dynamic random-access memory (DRAM) where reads or writes of bits can flip other bits. DRAM manufacturers add mitigations, but don’t disclose details, making it difficult for customers to evaluate their efficacy. We present a tool, based on active learning, that automatically infers parameter of Rowhammer mitigations against synthetic models of modern DRAM.