Search

Regulations increasingly call for various assurances from machine learning (ML) model providers about their training data, training process, and the behavior of resulting models during inference. For better transparency, companies (e.g., Huggingface and Google) have adopted model cards and datasheets which describe different properties of the training datasets and models. In the same vein, we introduce the notion of an inference card to describe the properties of a given inference (e.g., binding output to the model and its corresponding input). We collectively refer to these as ML property cards. A malicious model provider can include false information in ML property cards, raising a need for verifiable ML property cards. We show how to realized them using property attestation, technical mechanisms by which a prover (e.g., a model provider) can attest different ML properties during training and inference to a verifier (e.g., an auditor). However, prior attestation mechanisms based purely on cryptography are often narrowly focused (lacking versatility) and inefficient. There is a need to efficiently attest different types properties across the ML model training and inference pipeline. Recent developments make it possible to run and even train models inside hardware-assisted trusted execution environments (TEEs), which can provide highly efficient attestation. We propose Laminator, the first framework for verifiable ML property cards using hardware-assisted ML property attestations to efficiently furnish attestations for various ML properties for training and inference. It scales to multiple verifiers, and is independent of the model configuration.

Outsourced computation can put client data confidentiality at risk. Existing solutions are either inefficient or insufficiently secure: cryptographic techniques like fully-homomorphic encryption incur significant overheads, even with hardware assistance, while the complexity of hardware-assisted trusted execution environments has been exploited to leak secret data. Recent proposals such as BliMe and OISA show how dynamic information flow tracking (DIFT) enforced in hardware can protect client data efficiently. They are designed to protect CPU-only workloads. However, many outsourced computing applications, like machine learning, make extensive use of accelerators. We address this gap with Dolma, which applies DIFT to the Gemmini matrix multiplication accelerator, efficiently guaranteeing client data confidentiality, even in the presence of malicious/vulnerable software and side channel attacks on the server. We show that accelerators can allow DIFT logic optimizations that significantly reduce area overhead compared with general-purpose processor architectures. Dolma is integrated with the BliMe framework to achieve end-to-end security guarantees. We evaluate Dolma on an FPGA using a ResNet-50 DNN model and show that it incurs low overheads for large configurations ($4.4\%$, $16.7\%$, $16.5\%$ for performance, resource usage and power, respectively, with a 32x32 configuration).

Modern processors include high-performance cryptographic functionalities such as Intel's AES-NI and ARM's Pointer Authentication that allow programs to efficiently authenticate data held by the program. Pointer Authentication is already used to protect return addresses in recent Apple devices, but as yet these structures have seen little use for the protection of general program data. In this paper, we show how cryptographically-authenticated data structures can be used to protect against attacks based on memory corruption, and show how they can be efficiently realized using widely available hardware-assisted cryptographic mechanisms. We present realizations of secure stacks and queues with minimal overall performance overhead (3.4%-6.4% slowdown of the OpenCV core performance tests), and provide proofs of correctness., Comment: Presented at the 2022 IEEE Secure Development Conference. Copyright 2022 IEEE

Nonlinearity in many systems is heavily dependent on component variation and environmental factors such as temperature. This is often overcome by keeping signals close enough to the device's operating point that it appears approximately linear. But as the signal being measured becomes larger, the deviation from linearity increases, and the device's nonlinearity specification will be exceeded. This limits the range over which the device will produce directly useful measurements, often to far less than the device's safe range of operation.

Outsourced computing is widely used today. However, current approaches for protecting client data in outsourced computing fall short: use of cryptographic techniques like fully-homomorphic encryption incurs substantial costs, whereas use of hardware-assisted trusted execution environments has been shown to be vulnerable to run-time and side-channel attacks. We present BliMe, an architecture to realize efficient and secure outsourced computation. BliMe consists of a novel and minimal set of instruction set architecture (ISA) extensions implementing a taint-tracking policy to ensure the confidentiality of client data even in the presence of server vulnerabilities. To secure outsourced computation, the BliMe extensions can be used together with an attestable, fixed-function hardware security module (HSM) and an encryption engine that provides atomic decrypt-and-taint and encrypt-and-untaint operations. Clients rely on remote attestation and key agreement with the HSM to ensure that their data can be transferred securely to and from the encryption engine and will always be protected by BliMe's taint-tracking policy while at the server. We provide an RTL implementation BliMe-BOOM based on the BOOM RISC-V core. BliMe-BOOM requires no reduction in clock frequency relative to unmodified BOOM, and has minimal power ($\lt1.5\%$) and FPGA resource ($\leq9.0\%$) overheads. Various implementations of BliMe incur only moderate performance overhead ($8-25\%$). We also provide a machine-checked security proof of a simplified model ISA with BliMe extensions.

The Cattle Tracing System database is an online recording system for cattle births, deaths and between--herd movements in the United Kingdom. Although it has been thoroughly examined, the most recently reported movement analysis is from 2009. This article uses the database to construct weighted directed monthly movement networks for two distinct periods of time, 2004--2006 and 2015--2017, to quantify by how much the underlying structure of the network has changed. Substantial changes in network structure may influence policy--makers directly or may influence models built upon the network data, and these in turn could impact policy--makers and their assessment of risk. Four general network measures are used (total number of nodes with movements, movements, births and deaths), in conjunction with network metrics to describe each monthly network. Two updates of the database were examined to determine by how much the movement data stored for a particular time period had been cleansed between updates. Statistical models show that there is a statistically significant effect of the time period (2004--2006 vs 2015--2017) in the values of all network measures and six of nine network metrics. Changes in the sizes of both the Giant and Weakly Strongly Connected components predict reductions in the upper and lower bounds of the maximum epidemic size. Examination of the updates of the database show that there are differences in records between updates and therefore evidence of historical data changing between updates. Accurate modelling of disease spread through a network requires representative descriptions of the network. The authors recommend that where possible the most recent available data always be used for network modelling and that methods of network prediction be examined to mitigate for the time required for data to become available.

Autonomous collaborative networks of devices are rapidly emerging in numerous domains, such as self-driving cars, smart factories, critical infrastructure, and Internet of Things in general. Although autonomy and self-organization are highly desired properties, they increase vulnerability to attacks. Hence, autonomous networks need dependable mechanisms to detect malicious devices in order to prevent compromise of the entire network. However, current mechanisms to detect malicious devices either require a trusted central entity or scale poorly. In this paper, we present GrandDetAuto, the first scheme to identify malicious devices efficiently within large autonomous networks of collaborating entities. GrandDetAuto functions without relying on a central trusted entity, works reliably for very large networks of devices, and is adaptable to a wide range of application scenarios thanks to interchangeable components. Our scheme uses random elections to embed integrity validation schemes in distributed consensus, providing a solution supporting tens of thousands of devices. We implemented and evaluated a concrete instance of GrandDetAuto on a network of embedded devices and conducted large-scale network simulations with up to 100000 nodes. Our results show the effectiveness and efficiency of our scheme, revealing logarithmic growth in run-time and message complexity with increasing network size. Moreover, we provide an extensive evaluation of key parameters showing that GrandDetAuto is applicable to many scenarios with diverse requirements.

Consensus mechanisms used by popular distributed ledgers are highly scalable but notoriously inefficient. Byzantine fault tolerance (BFT) protocols are efficient but far less scalable. Speculative BFT protocols such as Zyzzyva and Zyzzyva5 are efficient and scalable but require a trade-off: Zyzzyva requires only $3f + 1$ replicas to tolerate $f$ faults, but even a single slow replica will make Zyzzyva fall back to more expensive non-speculative operation. Zyzzyva5 does not require a non-speculative fallback, but requires $5f + 1$ replicas in order to tolerate $f$ faults. BFT variants using hardware-assisted trusted components can tolerate a greater proportion of faults, but require that every replica have this hardware. We present SACZyzzyva, addressing these concerns: resilience to slow replicas and requiring only $3f + 1$ replicas, with only one replica needing an active monotonic counter at any given time. We experimentally evaluate our protocols, demonstrating low latency and high scalability. We prove that SACZyzzyva is optimally robust and that trusted components cannot increase fault tolerance unless they are present in greater than two-thirds of replicas., Comment: \copyright\ 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works