Browse wiki

From Navigators

Jump to: navigation, search
Abstract The Cloud Computing model has incentivized The Cloud Computing model has incentivized companies to outsource services to third-party providers. Service owners can use third-party computational, storage and network resources while avoiding the cost of acquiring an IT infrastructure. However, they have to rely on the trustworthiness of the third-party providers, who ultimately need to guarantee that the services run as intended. The fundamental security challenge is how to empower companies that own and outsource such services, or clients using them, to check service execution on the remote cloud platform. A promising approach is based on hardware-enforced isolation and attestation of the service execution. Assuming that hardware attacks are infeasible, this protects the service from other malicious software or untrusted system administrators. Also, it allows clients to check that the results were produced as intended. While this paradigm is well known, previous work does not scale with large code and data sizes, lacks generality both with respect to hardware (e.g., either uses Trusted Platform Modules, TPMs, or Intel SGX) and software (e.g., only supports MapReduce applications), and makes undesirable security tradeoffs (e.g., resorts to a large Trusted Computing base, or TCB, to run unmodified services, or a small TCB but with limited functionality). This thesis shows how to secure the execution of large-scale services efficiently and without these compromises. From the perspective of a client that sends a request and receives a response, trust can be established by verifying a small proof of correct execution that is attached to the result. On the remote provider’s platform, a small trusted computing base enables the secure execution of generic services composed of a large source code base and/or working on large data sets, using an abstraction layer that is implementable on diverse trusted hardware architectures. Our small TCB implements three orthogonal techniques that are the core contributions of this thesis. The first one targets the identification (and the execution) of only the part of code that is necessary to fulfill a client’s request. This allows an increase both in security and efficiency by leaving any code that is not required to run the service outside the execution environment. The second contribution enables terabyte-scale data processing by means of a secure in-memory data handling mechanism. This allows a service to retrieve data that is validated on access and before use. Notably, data I/O is performed using virtual memory mechanisms that do not require any system call from the trusted execution environment, thereby reducing the attack surface. The third contribution is a novel fully-passive secure replication scheme that is tolerant to software attacks. Fault-tolerance delivers availability guarantees to clients, while passive replication allows for computationally efficient processing. Interestingly, all of our techniques are based on the same abstraction layer of the trusted hardware. In addition, our implementation and experimental evaluation demonstrate the practicality of these approaches. rate the practicality of these approaches.
Advisor Nuno Ferreira Neves + , Peter Steenkiste +
Author Bruno Vavala +
Document Document for Publication-BrunoVavalaTese.pdf +
Key BrunoVavalaTese  +
Month jul  +
NumPubDate 2,017.07  +
ResearchLine Fault and Intrusion Tolerance in Open Distributed Systems (FIT) +
School Departamento de Informática, Faculdade de Ciências, Universidade de Lisboa  +
Title Secure Large-Scale Outsourced Services Founded on Trustworthy Code Executions  +
Type phdthesis  +
Year 2017  +
Has improper value forThis property is a special property in this wiki. Url  +
Categories Publication  +
Modification dateThis property is a special property in this wiki. 2 October 2018 15:53:35  +
hide properties that link here 
  No properties link to this page.


Enter the name of the page to start browsing from.
Personal tools
Navigators toolbox