On the challenges of energy efficiency, scalability and security for internet of things services
The frontier of the Internet is a highly volatile and rapidly evolving area with Internet of Things (IoT), mobile, and smart devices. Every day more and more devices are being deployed and connected to clouds in order to form distributed systems at any scale. Especially large-scale IoT systems make use of public cloud infrastructure for demanding IoT data processing. As a result of this continuous expansion, common network infrastructures and cloud data centers have to shoulder a greater part of expanding Internet traffic. Furthermore, with their growing number of features, these systems offer an increasingly large attack surface to cybercriminals and malicious users. Once deployed, IoT devices are constantly exposed to a wide range of threats. Unauthorized persons can gain physical access to devices or devices can be compromised remotely by software exploits and vulnerabilities over the network. The increasing availability of hardware-based trusted execution environments, ranging from low-power embedded devices to high-performance servers, could be a solution to these challenges. In particular the lack of trust towards cloud providers has limited the deployment of large-scale distributed systems for manipulating confidential and sensitive data. Public and private companies, as well as organizations or individuals, are taking major risks by exposing their assets and business interactions to the cloud. Under some circumstances, it can even be illegal to disclose certain data or documents to the public, such as medical records. Exploiting trusted execution environments enables dissemination and processing of confidential and sensitive data in the cloud without having to trust the cloud provider. Alternative solutions, such as homomorphic encryption, address the same trust issues but remain in their current state in terms of computational performance and energy consumption behind requirements of current workloads. Most importantly, hardware heterogeneity of trusted execution environments allows selecting platforms based on the energy requirements of software services. Hardware, firmware, and operating system support is necessary for enabling complex interactions between various components in order to provide fully functional trusted execution environments. Software frameworks abstract the underlying complexity of these interactions and provide user space programming interfaces for large-scale adaptation of trusted execution environments. Understanding how IoT services shielded by trusted execution environments are being affected in terms of energy efficiency, scalability and security remains mostly unclear. This knowledge is particularly important for developing secure and scalable IoT services that run on battery-powered devices as well as to extend the availability and uptime of services. Many IoT devices rely on battery power or renewable energy sources, such as solar energy through solar panels. Critical supply voltage levels are reached when the battery is nearly empty or if insufficient solar power is provided, allowing for locally or remotely deployed voltages glitches to compromise security guarantees of devices. Under these conditions the integrity and security of IoT credentials, data, services and the system as a whole are at stake. Therefore, not only security, but also energy levels have to be considered when designing such systems, since particularly in IoT, energy and security are strongly intertwined. Novel approaches need to be considered for IoT services deployed in public and private clouds for determining safe undervolting margins and detecting maliciously undervolted cloud instances. The current thesis presents a selection of security-relevant IoT services which are enhanced by trusted execution environments in order to provide better security guarantees. These IoT services include design, implementation, deployment, and evaluation of blockchain, key-value and secure storage, as well as network performance systems on low-power ARM-based hardware. Energy efficiency, scalability, and security compromises are quantified in small-scale deployments for the evaluation. The secure IoT services are complemented by two novel energy-related approaches. The first approach is motivating the use of generalized deduplication in IoT network protocols to reduce energy, transmission, as well as storage costs. Undermining the reliability of processors is explored in a second approach where an undervolted cloud infrastructure is considered in an attack. The primary interest of this thesis are low-power IoT devices using ARM TrustZone in combination with the OP-TEE framework as trusted execution environment.
Read more