In this paper we tackle the open paradoxical challenge of FPGA-accelerated cloud computing: On one hand, clients aim to secure their Intellectual Property (IP) by encrypting their configuration bitstreams prior to uploading them to the cloud. On the other hand, cloud service providers disallow the use of encrypted bitstreams to mitigate rogue configurations from damaging or disabling the FPGA. Instead, cloud providers require a verifiable check on the hardware design that is intended to run on a cloud FPGA at the netlist-level before generating the bitstream and loading it onto the FPGA, therefore, contradicting the IP protection requirement of clients. Currently, there exist no practical solution that can adequately address this challenge.We present the first practical solution that, under reasonable trust assumptions, satisfies the IP protection requirement of the client and provides a bitstream sanity check to the cloud provider. Our proof-of-concept implementation uses existing tools and commodity hardware. It is based on a trusted FPGA shell that utilizes less than 1% of the FPGA resources on a Xilinx VCU118 evaluation board, and an Intel SGX machine running the design checks on the client bitstream.

Performing machine learning tasks in mobile applications yields a challenging conflict of interest: highly sensitive client information (e.g., speech data) should remain private while also the intellectual property of service providers (e.g., model parameters) must be protected. Cryptographic techniques offer secure solutions for this, but have an unacceptable overhead and moreover require frequent network interaction.In this work, we design a practically efficient hardware-based solution. Specifically, we build OFFLINE MODEL GUARD (OMG) to enable privacy-preserving machine learning on the predominant mobile computing platform ARM—even in offline scenarios. By leveraging a trusted execution environment for strict hardware-enforced isolation from other system components, OMG guarantees privacy of client data, secrecy of provided models, and integrity of processing algorithms. Our prototype implementation on an ARM HiKey 960 development board performs privacy-preserving keyword recognition using TensorFlow Lite for Microcontrollers in real time.

Recent research has demonstrated that Intel's SGX is vulnerable to software-based side-channel attacks. In a common attack, the adversary monitors CPU caches to infer secret-dependent data accesses patterns. Known defenses have major limitations, as they require either error-prone developer assistance, incur extremely high runtime overhead, or prevent only specific attacks.

With the advent of smart-home devices providing voice-based interfaces, such as Amazon Alexa or Apple Siri, voice data is constantly transferred to cloud services for automated speech recognition or speaker verification.While this development enables intriguing new applications, it also poses significant risks: Voice data is highly sensitive since it contains biometric information of the speaker as well as the spoken words.This data may be abused if not protected properly, thus the security and privacy of billions of end-users is at stake.We tackle this challenge by proposing an architecture, dubbed VoiceGuard, that efficiently protects the speech processing task inside a trusted execution environment (TEE).Our solution preserves the privacy of users while at the same time it does not require the service provider to reveal model parameters.Our architecture can be extended to enable user-specific models, such as feature transformations (including fMLLR), i-vectors, or model transformations (e.g., custom output layers).It also generalizes to secure on-premise solutions, allowing vendors to securely ship their models to customers.We provide a proof-of-concept implementation and evaluate it on the Resource Management and WSJ speech recognition tasks isolated with Intel SGX, a widely available TEE implementation, demonstrating even real time processing capabilities.

Memory-corruption vulnerabilities pose a serious threat to modern computer security. Attackers exploit these vulnerabilities to manipulate code and data of vulnerable applications to generate malicious behavior by means of code-injection and code-reuse attacks. Researchers already demonstrated the power of data-only attacks by disclosing secret data such as cryptographic keys in the past. A large body of literature has investigated defenses against code-injection, code-reuse, and data-only attacks. Unfortunately, most of these defenses are tailored towards statically generated code and their adaption to dynamic code comes with the price of security or performance penalties. However, many common applications, like browsers and document viewers, embed just-in-time compilers to generate dynamic code. The contribution of this paper is twofold: first, we propose a generic data-only attack against JIT compilers, dubbed DOJITA. In contrast to previous data-only attacks that aimed at disclosing secret data, DOJITA enables arbitrary code-execution. Second, we propose JITGuard, a novel defense to mitigate code-injection, code-reuse, and data-only attacks against just-in-time compilers (including DOJITA). JITGuard utilizes Intel's Software Guard Extensions (SGX) to provide a secure environment for emitting the dynamic code to a secret region, which is only known to the JIT compiler, and hence, inaccessible to the attacker. Our proposal is the first solution leveraging SGX to protect the security critical JIT compiler operations, and tackles a number of difficult challenges. As proof of concept we implemented JITGuard for Firefox's JIT compiler SpiderMonkey. Our evaluation shows reasonable overhead of 9.8% for common benchmarks.

The emergence of numerous new manufacturers producing devices for the Internet-of-Things (IoT) has given rise to new security concerns. Many IoT devices exhibit security flaws making them vulnerable for attacks and manufacturers have difficulties in providing appropriate security patches to their products in a timely and user-friendly manner. In this paper, we present our implementation of IoT Sentinel, which is a system aimed at protecting the user's network from vulnerable IoT devices. IoT Sentinel automatically identifies vulnerable devices when they are first introduced to the network and enforces appropriate traffic filtering rules to protect other devices from the threats originating from the vulnerable devices.

Abstract Tor is a well-known anonymous communication system used by millions of users, including journalists and civil rights activists all over the world. The Tor Browser gives non-technical users an easy way to access the Tor Network. However, many government organizations are actively trying to compromise Tor not only in regions with repressive regimes but also in the free world, as the recent FBI incidents clearly demonstrate. Exploiting software vulnerabilities in general, and browser vulnerabilities in particular, constitutes a clear and present threat to the Tor software. The Tor Browser shares a large part of its attack surface with the Firefox browser. Therefore, Firefox vulnerabilities (even patched ones) are highly valuable to attackers trying to monitor users of the Tor Browser. In this paper, we present selfrando-an enhanced and practical load-time randomization technique for the Tor Browser that defends against exploits, such as the one FBI allegedly used against Tor users. Our solution significantly improves security over standard address space layout randomization (ASLR) techniques currently used by Firefox and other mainstream browsers. Moreover, we collaborated closely with the Tor Project to ensure that selfrando is fully compatible with AddressSanitizer (ASan), a compiler feature to detect memory corruption. ASan is used in a hardened version of Tor Browser for test purposes. The Tor Project decided to include our solution in the hardened releases of the Tor Browser, which is currently undergoing field testing.