Stefan Saroiu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Stefan Saroiu has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Alec Wolman
Microsoft, Redmond, Washington, United States
3 shared publications - Bohdan Borysei
University of Toronto, Toronto, ON, Canada
2 shared publications - Eyal de Lara
University of Toronto, Toronto, Canada
2 shared publications - Ankit Verma
Microsoft, Cambridge, United Kingdom
1 shared publication - Anuj Kalia
Microsoft, Redmond, USA
1 shared publication - Baris Kasikci
University of Washington, Seattle, WA, USA
1 shared publication
Latest Publications
- Open Access
- 25 July 2025
- Published by Association for Computing Machinery (ACM) in Communications of the ACM
- Vol. 68 (8) , 81
Defending Data from SRAM-Based Attacks
- 25 July 2025
- Published by Association for Computing Machinery (ACM) in Communications of the ACM
- Vol. 68 (8) , 81
Open AccessDefending Data from SRAM-Based Attacks
- Open Access
- 28 February 2024
- Published by Association for Computing Machinery (ACM)
- p. 146
Poster: Towards Multi-Stakeholder Clouds
Cloud providers are embracing edge computing by developing multi-stakeholder clouds (MClouds), in which the infrastructure is shared among cloud providers, hosting site operators, and, sometimes, system integrators (Figure 1). The emergence of the MCloud comes with a new set of systems challenges stemming from its multi-stakeholder nature. In public clouds, the responsibility for infrastructure lies with the cloud provider. The cloud provider owns, manages, and operates all their servers and switches inside their datacenter. However, in the case of MClouds, the division of responsibilities becomes less clear. Should the cloud provider still handle tasks like powering servers, installing firmware updates, and monitoring server health, or is this now the responsibility of the hosting site's administrator team?
- 28 February 2024
- Published by Association for Computing Machinery (ACM)
- p. 146
Open AccessPoster: Towards Multi-Stakeholder Clouds
Cloud providers are embracing edge computing by developing multi-stakeholder clouds (MClouds), in which the infrastructure is shared among cloud providers, hosting site operators, and, sometimes, system integrators (Figure 1). The emergence of the MCloud comes with a new set of systems challenges stemming from its multi-stakeholder nature. In public clouds, the responsibility for infrastructure lies with the cloud provider. The cloud provider owns, manages, and operates all their servers and switches inside their datacenter. However, in the case of MClouds, the division of responsibilities becomes less clear. Should the cloud provider still handle tasks like powering servers, installing firmware updates, and monitoring server health, or is this now the responsibility of the hosting site's administrator team?
- Open Access
- 28 February 2024
- Published by Association for Computing Machinery (ACM)
- p. 47-52
Towards Multi-Stakeholder Clouds
Cloud providers are embracing edge computing by developing multi-stakeholder clouds (MClouds), in which the infrastructure is shared among cloud providers, hosting site operators, and, sometimes, system integrators. This form of sharing creates a new set of challenges around infrastructure management. Unfortunately, traditional resource management methods based on partitioning and virtualization do not fit well the needs of infrastructure management in a multi-stakeholder scenario. Our paper describes these new set of challenges, and puts forward a new authorization framework based on two-person control (P2C).
- 28 February 2024
- Published by Association for Computing Machinery (ACM)
- p. 47-52
Open AccessTowards Multi-Stakeholder Clouds
Cloud providers are embracing edge computing by developing multi-stakeholder clouds (MClouds), in which the infrastructure is shared among cloud providers, hosting site operators, and, sometimes, system integrators. This form of sharing creates a new set of challenges around infrastructure management. Unfortunately, traditional resource management methods based on partitioning and virtualization do not fit well the needs of infrastructure management in a multi-stakeholder scenario. Our paper describes these new set of challenges, and puts forward a new authorization framework based on two-person control (P2C).
- Open Access
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 417-433
Siloz: Leveraging DRAM Isolation Domains to Prevent Inter-VM Rowhammer
Today's cloud DRAM lacks strong isolation primitives, highlighted by Rowhammer bit flips. Rowhammer poses an increasing threat to cloud security/reliability, given (1) DRAM activation rates in commodity and malicious workloads already exceed Rowhammer thresholds, and (2) thresholds are decreasing in newer DRAM. Deployed hardware mitigations remain vulnerable, turning cloud providers toward software defenses. However, existing defenses incur high performance or memory overhead or contain significant protection gaps. Accordingly, we introduce Siloz, a hypervisor that uses subarray groups as DRAM isolation domains to enable efficient protection against inter-VM Rowhammer. Siloz exploits the insights that (a) Rowhammer can only flip bits in DRAM rows located in the same subarray---not across subarrays---and (b) VMs can be isolated to groups of subarrays without sacrificing bank-level parallelism, a key component of DRAM performance. Siloz thus prevents inter-VM bit flips by placing each VM's and the host's data into private subarray groups. To additionally ensure that a VM cannot escape its provisioned subarray group(s), Siloz provides integrity protection for extended page tables (EPTs). We show that Siloz's implementation has negligible effect on average performance across various cloud workloads, SPEC CPU 2017, and PARSEC 3.0 (within ±0.5% of baseline Linux/KVM).
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 417-433
Open AccessSiloz: Leveraging DRAM Isolation Domains to Prevent Inter-VM Rowhammer
Today's cloud DRAM lacks strong isolation primitives, highlighted by Rowhammer bit flips. Rowhammer poses an increasing threat to cloud security/reliability, given (1) DRAM activation rates in commodity and malicious workloads already exceed Rowhammer thresholds, and (2) thresholds are decreasing in newer DRAM. Deployed hardware mitigations remain vulnerable, turning cloud providers toward software defenses. However, existing defenses incur high performance or memory overhead or contain significant protection gaps. Accordingly, we introduce Siloz, a hypervisor that uses subarray groups as DRAM isolation domains to enable efficient protection against inter-VM Rowhammer. Siloz exploits the insights that (a) Rowhammer can only flip bits in DRAM rows located in the same subarray---not across subarrays---and (b) VMs can be isolated to groups of subarrays without sacrificing bank-level parallelism, a key component of DRAM performance. Siloz thus prevents inter-VM bit flips by placing each VM's and the host's data into private subarray groups. To additionally ensure that a VM cannot escape its provisioned subarray group(s), Siloz provides integrity protection for extended page tables (EPTs). We show that Siloz's implementation has negligible effect on average performance across various cloud workloads, SPEC CPU 2017, and PARSEC 3.0 (within ±0.5% of baseline Linux/KVM).
- Open Access
- 2 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-3
Accelerating Open RAN Research Through an Enterprise-scale 5G Testbed
Open RAN is an emerging paradigm in mobile networks where the Radio Access Network (RAN) functions are disaggregated and virtualized on commodity servers. Despite the importance of Open RAN research, existing platforms often lack the fidelity and stability required to address a wide range of research problems. In response to this limitation, we have developed an enterprise-scale Open RAN testbed aimed at conducting state-of-the-art research in key areas that have received limited attention due to the lack of suitable platforms. In this poster, we provide an overview of the testbed we have created and examples of the research it has enabled, with the hope of catalyzing future open RAN research and innovation.
- 2 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-3
Open AccessAccelerating Open RAN Research Through an Enterprise-scale 5G Testbed
Open RAN is an emerging paradigm in mobile networks where the Radio Access Network (RAN) functions are disaggregated and virtualized on commodity servers. Despite the importance of Open RAN research, existing platforms often lack the fidelity and stability required to address a wide range of research problems. In response to this limitation, we have developed an enterprise-scale Open RAN testbed aimed at conducting state-of-the-art research in key areas that have received limited attention due to the lack of suitable platforms. In this poster, we provide an overview of the testbed we have created and examples of the research it has enabled, with the hope of catalyzing future open RAN research and innovation.