Publications

In the last five years, FPGA presence in the cloud has gone from near zero (except for deeply embedded devices) …

In this article, we survey existing academic and commercial efforts to provide Field-Programmable Gate Array (FPGA) acceleration in datacenters and …

Host-FPGA connectivity is critical for enabling a vast number of FPGA use cases in data centers, edge, and IoT. This …

High Level Synthesis (HLS) offers a possible programmability solution for FPGAs by automatically compiling CPU codes to custom hardware configurations, …

Fully Homomorphic Encryption (FHE) offers protection to private data on third-party cloud servers by allowing computations on the data in encrypted form. However, to support general-purpose encrypted computations, all existing FHE schemes require an expensive operation known as “bootstrapping”. Unfortunately, the computation cost and the memory bandwidth required for bootstrapping add significant overhead to FHE-based computations, limiting the practical use of FHE.

In this work, we propose FAB, an FPGA-based accelerator for bootstrappable FHE. Prior FPGA-based FHE accelerators have proposed hardware acceleration of basic FHE primitives for impractical parameter sets without support for bootstrapping.

Analytical database systems are typically designed to use a column first data layout to access only the desired fields. On the other hand, storing data row-first works great for accessing, inserting, or updating entire rows. Transforming rows to columns at runtime is expensive, hence, many analytical systems ingest data in row-first form and transform it in the background to columns to facilitate future analytical queries. How will this design change if we can always efficiently access only the desired set of columns?

To address this question, we present a radically new approach to data transformation from rows to columns. We build upon recent advancements in embedded platforms with re-programmable logic to design native in-memory access on rows and columns.

This paper introduces GCNSplit, a streaming graph partitioning framework capable of handling unbounded streams with bounded state requirements. We frame partitioning as a classification problem and we employ an unsupervised model whose loss function minimizes edge-cuts. GCNSplit leverages an inductive graph convolutional network (GCN) to embed graph characteristics into a low-dimensional space and assign edges to partitions in an online manner. We evaluate GCNSplit with real-world graph datasets of various sizes and domains. Our results demonstrate that GCNSplit provides high-throughput, top-quality partitioning, and successfully leverages data parallelism. It achieves a throughput of 430K edges/s on a real-world graph of 1.6B edges using a bounded 147KB-sized model, contrary to the state-of-the-art HDRF algorithm that requires > 116GB in-memory state. With a well-balanced normalized load of 1.01, GCNSplit achieves a replication factor on par with HDRF, showcasing high partitioning quality while storing three orders of magnitude smaller partitioning state. Owing to the power of GCNs, we show that GCNSplit can generalize to entirely unseen graphs while outperforming the state-of-the-art stream partitioners in some cases.

We present the first performance evaluation study of model serving integration tools in stream processing frameworks. Using Apache Flink as a representative stream processing system, we evaluate alternative Deep Learning serving pipelines for image classification. Our performance evaluation considers both the case of embedded use of Machine Learning libraries within stream tasks and that of external serving via Remote Procedure Calls. The results indicate superior throughput and scalability for pipelines that make use of embedded libraries to serve pre-trained models. Whereas, latency can vary across strategies, with external serving even achieving lower latency when network conditions are optimal due to better specialized use of underlying hardware. We discuss our findings and provide further motivating arguments towards research in the area of ML-native data streaming engines in the future.

We present a vision for the automatic optimisation of distributed stream processing programs. StrIoT — a distributed stream-processing framework built using purely-functional programming — enables a set of validated logical optimisation rules to generate a set of possible deployment plans. A cost model then filters and ranks the plans before the best is automatically deployed across the cloud and edge devices. We describe StrIoT’s functional operators for writing stream-processing programs; the design, implementation and performance of StrIoT’s logical optimiser; and the cost model, which filters and ranks re-written programs and deployment plans in terms of two non-functional requirements: bandwidth and cost. The StrIoT vision is being explored through an open-source proof-of-concept implementation. We present our initial results with a motivating example before outlining the success criteria for future work in this area.

Applications that process streams of events generated by sensors are important in a wide range of domains. It is now common to distribute stream processing across edge devices and the cloud. This exploits processing power near the sensors, reducing the load on the cloud and often the required network bandwidth. In this paper we focus on one challenge in distributed stream processing: automatically adapting the partitioning of the processing between the edge and the cloud without a loss of service. An example is when the event arrival rate increases and the edge processor can no longer meet performance requirements. Re-partitioning without loss of service involves moving computations between the edge and the cloud while events are still being processed. In this paper we describe StrIoT – a stream processing system that supports automatic re-partitioning. It is based on a set of functional stream operators, and the paper describes how the run-time system can automatically adapt applications that use them. A key feature is support for the fission and fusion of operators during adaptations. Performance evaluation shows that StrIoT can move parts of a stream processing application between the cloud and edge with only a low, temporary impact on performance.

In recent years, we have witnessed a new kind of DDoS attack, the burst attack(Chai, 2013; Dahan, 2018), where the …

Auto-scaling mechanisms are an important line of defense against Distributed Denial of Service (DDoS) in the cloud. Using auto-scaling, machines …

Paper presented at 17th Swedish National Computer Networking Workshop (SNCNW 2022), KTH, Stockholm, June 16-17, 2022 Abstract: Network latency plays …

The Linux eXpress Data Path, or XDP, has found numerous uses in the industry, such as mitigating DoS attacks, load-balancers, and intrusion prevention systems. XDP provides a high-performance programmable network data path using the BPF framework and al- lows programmers to process packets early out of the driver. While XDP excels in forwarding packets, it currently has no mechanism for queueing or reordering packets and cannot implement traffic scheduling policies. In this paper, we present our ongoing work to address this challenge. We have designed a programmable packet scheduling extension for the XDP framework using recently pro- posed schemes for programmable queues. This extension allows programmers to define their packet schedulers using BPF while benefiting from the XDP fast data path.

Abstract Industrial microservice architectures vary so wildly in their characteristics, such as size or communication method, that comparing systems is …

Abstract: Log-Structured Merge trees (LSM trees) are increasingly used as the storage engines behind several data systems, frequently deployed in …

Abstract: Effectively utilizing the vast amounts of ego-centric navigation data that is freely available on the internet can advance generalized …

Block-based programming is a popular approach to teaching introductory programming. Block-based programming often works in the context of microworlds, where students solve specific puzzles. It is used, for example, within the Hour of Code event, which targets millions of students.

On the internet, images are no longer static; they have become dynamic content. Thanks to the availability of smartphones with …

The proliferation of multi-core, accelerator-enabled embedded systems has introduced new opportunities to consolidate real-time systems of increasing complexity. But the …

FPGAs have typically achieved high speedups for 3D Fast Fourier Transforms (FFTs) due to the presence of hard floating point …

Improved support for OpenCL has been an important step towards the mainstream adoption of FPGAs as compute resources. Current research …

FPGAs have been demonstrated to be capable of very high performance, especially power-performance, but generally at the cost of hand-tuned …

As modern Data Center workloads become increasingly complex, constrained, and critical, mainstream CPU-centric computing has had ever more difficulty in …

Multi-Party Computation (MPC) is a technique enabling data from several sources to be used in a secure computation revealing only …

Kariz is a new architecture for caching data from data-lakes accessed, potentially concurrently, by multiple analytic platforms. It integrates rich …

“Many analytics computations are dominated by iterative processing stages, executed until a convergence condition is met. To accelerate such workloads …

“In recent years, heterogeneous computing has emerged as the vital way to increase computers’ performance and energy efficiency by combining …

The ever-growing demand for more memory capacity from applications has always been a challenging factor in computer architecture. The advent …

Introduction – why stream processing? It is commonly known that we are living in the era of Big Data. At …

We analyse whether the smartcards of the JavaCard platform correctly validate primality of domain parameters.

We present our discovery of a group of side-channel vulnerabilities in implementations of the ECDSA signature algorithm in a widely …

This report describes the results of testing of authentication methods and applications for smartphones, including the methodology that was used. …

Cyber-security protection of critical systems is one of the major challenges of today. Although the attacks typically originate from attackers …

The evaluation of student models involves many methodological decisions, e.g., the choice of performance metric, data filtering, and cross-validation setting. Such issues may seem like technical details, and they do not get much attention in published research.

Secure Multi-Party Computation (MPC) allows mutually distrusting parties to run joint computations without revealing private data. Current MPC algorithms scale …

Web browsers are increasingly used as middleware platforms offering a central access point for service provision. Using backend containerization, RESTful …

Existing bare-metal cloud services that provide users with physical servers have a number of serious disadvantages over their virtual alternatives, …

To get good performance for data stored in Object storage services like S3, data analysis clusters need to cache data …

Bolted is a new architecture for a bare metal cloud with the goal of providing security-sensitive customers of a cloud …

SecCloud is a new architecture for bare-metal clouds that enables tenants to control tradeoffs between security, price, and performance. It …

The received wisdom suggests that Unix’s unusual combination of fork() and exec() for process creation was an inspired design. In …

Unikernels have demonstrated enormous advantages over Linux in many important domains, causing some to propose that the days of Linux’s …

Modern FaaS systems perform well in the case of repeat executions when function working sets stay small. However, these platforms …

Current caching methods for improving the performance of big-data jobs assume high (e.g., full bi-section) bandwidth; however many enterprise data …

This paper presents a system-level method for achieving the rapid deployment and high-density caching of serverless functions in a FaaS …

Complex event processing (CEP) is widely employed to detect occurrences of predefined combinations (patterns) of events in massive data streams.

Complex event processing (CEP) is a prominent technology used in many modern applications for monitoring and tracking events of interest in massive data streams.

Additive Factors Model is a widely used student model, which is primarily used for refining knowledge component models (Q-matrices).

Continuous verification and security analysis of software systems are of paramount importance to many organizations. The state-of-the-art for such operations …

Software applications that employ secure multi-party computation (MPC) can empower individuals and organizations to benefit from privacy-preserving data analyses when …

Secure multi-party computation (MPC) enables joint computation over private data sets contributed by multiple entities. Using MPC in real-world software …

Additive Factors Model is a widely used student model, which is primarily used for refining knowledge component models (Q-matrices).

In this paper, we present a novel framework for real-timemulti-pattern complex event processing.

An important step toward adoption of formal methods in software development is support for mainstream programming languages. Unfortunately, these languages …

In model checking of real-life C and C++ programs, both search efficiency and counterexample readability are very important. In this …

We present a new release of the parallel and distributed LTL model checker DiVinE. The major improvement in this new release …

Model checking became a standard method of analyzing complex systems in many application domains. No doubt, a number of applications …

We present a new design and a C++ implementation of a high-performance, cache-efficient hash table suitable for use in implementation …

In this paper, we present substantial improvements in efficiency of explicit-state LTL model checking of C & C++ programs, building …

In this paper, we present substantial improvements in efficiency of explicit-state LTL model checking of C & C++ programs, building …

We present an extension of the DIVINE software model checker to support programs with exception handling. The extension consists of …

Abstract: The fourth version of the DIVINE model checker provides a modular platform for verification of real-world programs. It is …

Data races are among the most difficult software bugs to discover. They arise from multiple threads accessing the same memory …

Many full disk encryption applications rely on a strong password-based key derivation function to process a passphrase. This article defines …

Full disk encryption (FDE) is a common way to prevent unauthorized use of data by encrypting the whole storage device. …

Full Disk Encryption (FDE) provides confidentiality of a data-at-rest stored on persistent devices like disk or solid state drives (SSD). …

Research cooperation of CRoCS, Faculty of Informatics, Masaryk University and Red Hat. Authors: Martin Ukrop and Vashek Matyas Conference: RSA Cryptographers’ Track 2018 Paper details: Pge …

Many starting teachers of computer science have great professional skill but often lack pedagogical training. Since providing expert mentorship directly …

Flawed TLS certificates are not uncommon on the Internet. While they signal a potential issue, in most cases they have benign causes (e.g., misconfiguration or even deliberate deployment). This adds fuzziness to the decision on whether to trust a connection or not. Little is known about perceptions of flawed certificates by IT professionals, even though their decisions impact high numbers of end users. Moreover, it is unclear how much does the content of error messages and documentation influence these perceptions.

Programmable packet processing is increasingly implemented using kernel bypass  techniques, where a userspace application takes complete control of the  networking …

In this paper, we elaborate an abstract domain for C strings, that is, null-terminated arrays of characters. We describe the abstract semantics of basic string operations and prove their soundness with regards to previously established concrete semantics of those operations.

DIVINE is an LLVM-based verification tool focusing on anal- ysis of real-world C and C++ programs. Such programs often interact with their environment, for example via inputs from users or network. When these programs are analyzed, it is desirable that the verification tool can deal with inputs symbolically and analyze runs for all inputs. In DIVINE, it is now possible to deal with input data via symbolic computa- tion instrumented into the original program at the level of LLVM bitcode.

Symbolic computation is an important approach in automated program analysis. Most state-of-the-art tools perform symbolic computation as interpreters and directly maintain symbolic data. In this paper, we show that it is feasible, and in fact practical, to use a compiler-based strategy instead.