Abstract
The increasing level of sophistication of cyber attacks which are employing cross-cutting strategies that leverage multi-domain attack surfaces, including but not limited to, software defined networking poisoning, biasing of machine learning models to suppress detection, exploiting software (development), and leveraging system design deficiencies. While current defensive solutions exist, they only partially address multi-domain and multi-stage attacks, thus rendering them ineffective to counter the upcoming generation of attacks. More specifically, we argue that a disruption is needed to approach separated knowledge domains, namely Intrusion Tolerant systems, cybersecurity, and machine learning. We argue that current solutions tend to address different concerns/facets of overlapping issues and they tend to make strong assumptions of supporting infrastructure, e.g., assuming that event probes/metrics are not compromised. To address these issues, we present Skynet, a platform that acts as a secure overseer that merges traditional roles of SIEMs with conventional orchestrators while being rooted on the fundamentals introduced by previous generations of intrusion tolerant systems. Our goal is to provide an open-source intrusion tolerant platform that can dynamically adapt to known and unknown security threats in order to reduce potential vulnerability windows.

Abstract
Byzantine Fault tolerant (BFT) protocols are implemented to guarantee the correct system/application behavior even in the presence of arbitrary faults (i.e., Byzantine faults). Byzantine Fault tolerant State Machine Replication (BFT-SMR) is a known software solution for masking arbitrary faults and malicious attacks (Liu et al., 2020). In this survey, we present and discuss relevant BFT-SMR protocols, focusing on deterministic and probabilistic approaches. The main purpose of this paper is to discuss the characteristics of proposed works for each approach, as well as identify the trade-offs for each different approach.& COPY; 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )

Abstract
Many real-world datasets have an underlying dynamic graph structure, where entities and their interactions evolve over time. Machine learning models should consider these dynamics in order to harness their full potential in downstream tasks. Previous approaches for graph representation learning have focused on either sampling khop neighborhoods, akin to breadth-first search, or random walks, akin to depth-first search. However, these methods are computationally expensive and unsuitable for real-time, low-latency inference on dynamic graphs. To overcome these limitations, we propose graph-sprints a general purpose feature extraction framework for continuous-time-dynamic-graphs (CTDGs) that has low latency and is competitive with state-of-the-art, higher latency models. To achieve this, a streaming, low latency approximation to the random-walk based features is proposed. In our framework, time-aware node embeddings summarizing multi-hop information are computed using only single-hop operations on the incoming edges. We evaluate our proposed approach on three open-source datasets and two in-house datasets, and compare with three state-of-the-art algorithms (TGN-attn, TGN-ID, Jodie). We demonstrate that our graph-sprints features, combined with a machine learning classifier, achieve competitive performance (outperforming all baselines for the node classification tasks in five datasets). Simultaneously, graphsprints significantly reduce inference latencies, achieving close to an order of magnitude speed-up in our experimental setting.

Abstract
Machine learning methods to aid defence systems in detecting malicious activity typically rely on labelled data. In some domains, such labelled data is unavailable or incomplete. In practice this can lead to low detection rates and high false positive rates, which characterise for example anti-money laundering systems. In fact, it is estimated that 1.7-4 trillion euros are laundered annually and go undetected. We propose The GANfather, a method to generate samples with properties of malicious activity, without label requirements. We propose to reward the generation of malicious samples by introducing an extra objective to the typical Generative Adversarial Networks (GANs) loss. Ultimately, our goal is to enhance the detection of illicit activity using the discriminator network as a novel and robust defence system. Optionally, we may encourage the generator to bypass pre-existing detection systems. This setup then reveals defensive weaknesses for the discriminator to correct. We evaluate our method in two real-world use cases, money laundering and recommendation systems. In the former, our method moves cumulative amounts close to 350 thousand dollars through a network of accounts without being detected by an existing system. In the latter, we recommend the target item to a broad user base with as few as 30 synthetic attackers. In both cases, we train a new defence system to capture the synthetic attacks.

Abstract

Abstract
Many complex systems exist in the physical world and therefore can be modeled by networks in which their nodes and edges are embedded in space. However, classical network motifs only use purely topological information and disregard other features. In this paper we introduce a novel and general subgraph abstraction that incorporates spatial information, therefore enriching its characterization power. Moreover, we describe and implement a method to compute and count our spatial subgraphs in any given network. We also provide initial experimental results by using our methodology to produce spatial fingerprints of real road networks, showcasing its discrimination power and how it captures more than just simple topology.