Showing 1–17 of 17 results for author: Penzkofer, A

MultiMediate'24: Multi-Domain Engagement Estimation

Authors: Philipp Müller, Michal Balazia, Tobias Baur, Michael Dietz, Alexander Heimerl, Anna Penzkofer, Dominik Schiller, François Brémond, Jan Alexandersson, Elisabeth André, Andreas Bulling

Abstract: Estimating the momentary level of participant's engagement is an important prerequisite for assistive systems that support human interactions. Previous work has addressed this task in within-domain evaluation scenarios, i.e. training and testing on the same dataset. This is in contrast to real-life scenarios where domain shifts between training and testing data frequently occur. With MultiMediate'… ▽ More Estimating the momentary level of participant's engagement is an important prerequisite for assistive systems that support human interactions. Previous work has addressed this task in within-domain evaluation scenarios, i.e. training and testing on the same dataset. This is in contrast to real-life scenarios where domain shifts between training and testing data frequently occur. With MultiMediate'24, we present the first challenge addressing multi-domain engagement estimation. As training data, we utilise the NOXI database of dyadic novice-expert interactions. In addition to within-domain test data, we add two new test domains. First, we introduce recordings following the NOXI protocol but covering languages that are not present in the NOXI training data. Second, we collected novel engagement annotations on the MPIIGroupInteraction dataset which consists of group discussions between three to four people. In this way, MultiMediate'24 evaluates the ability of approaches to generalise across factors such as language and cultural background, group size, task, and screen-mediated vs. face-to-face interaction. This paper describes the MultiMediate'24 challenge and presents baseline results. In addition, we discuss selected challenge solutions. △ Less

Submitted 29 August, 2024; originally announced August 2024.

Comments: arXiv admin note: text overlap with arXiv:2308.08256

The Overcooked Generalisation Challenge

Authors: Constantin Ruhdorfer, Matteo Bortoletto, Anna Penzkofer, Andreas Bulling

Abstract: We introduce the Overcooked Generalisation Challenge (OGC) - the first benchmark to study agents' zero-shot cooperation abilities when faced with novel partners and levels in the Overcooked-AI environment. This perspective starkly contrasts a large body of previous work that has trained and evaluated cooperating agents only on the same level, failing to capture generalisation abilities required fo… ▽ More We introduce the Overcooked Generalisation Challenge (OGC) - the first benchmark to study agents' zero-shot cooperation abilities when faced with novel partners and levels in the Overcooked-AI environment. This perspective starkly contrasts a large body of previous work that has trained and evaluated cooperating agents only on the same level, failing to capture generalisation abilities required for real-world human-AI cooperation. Our challenge interfaces with state-of-the-art dual curriculum design (DCD) methods to generate auto-curricula for training general agents in Overcooked. It is the first cooperative multi-agent environment specially designed for DCD methods and, consequently, the first benchmarked with state-of-the-art methods. It is fully GPU-accelerated, built on the DCD benchmark suite minimax, and freely available under an open-source license: https://git.hcics.simtech.uni-stuttgart.de/public-projects/OGC. We show that current DCD algorithms struggle to produce useful policies in this novel challenge, even if combined with recent network architectures that were designed for scalability and generalisability. The OGC pushes the boundaries of real-world human-AI cooperation by enabling the research community to study the impact of generalisation on cooperating agents. △ Less

Submitted 3 April, 2025; v1 submitted 25 June, 2024; originally announced June 2024.

Comments: 26 pages

VSA4VQA: Scaling a Vector Symbolic Architecture to Visual Question Answering on Natural Images

Authors: Anna Penzkofer, Lei Shi, Andreas Bulling

Abstract: While Vector Symbolic Architectures (VSAs) are promising for modelling spatial cognition, their application is currently limited to artificially generated images and simple spatial queries. We propose VSA4VQA - a novel 4D implementation of VSAs that implements a mental representation of natural images for the challenging task of Visual Question Answering (VQA). VSA4VQA is the first model to scale… ▽ More While Vector Symbolic Architectures (VSAs) are promising for modelling spatial cognition, their application is currently limited to artificially generated images and simple spatial queries. We propose VSA4VQA - a novel 4D implementation of VSAs that implements a mental representation of natural images for the challenging task of Visual Question Answering (VQA). VSA4VQA is the first model to scale a VSA to complex spatial queries. Our method is based on the Semantic Pointer Architecture (SPA) to encode objects in a hyperdimensional vector space. To encode natural images, we extend the SPA to include dimensions for object's width and height in addition to their spatial location. To perform spatial queries we further introduce learned spatial query masks and integrate a pre-trained vision-language model for answering attribute-related questions. We evaluate our method on the GQA benchmark dataset and show that it can effectively encode natural images, achieving competitive performance to state-of-the-art deep learning methods for zero-shot VQA. △ Less

Submitted 6 May, 2024; originally announced May 2024.

Comments: To be published in the Proceedings of the Annual Meeting of the Cognitive Science Society (CogSci'24)

Int-HRL: Towards Intention-based Hierarchical Reinforcement Learning

Authors: Anna Penzkofer, Simon Schaefer, Florian Strohm, Mihai Bâce, Stefan Leutenegger, Andreas Bulling

Abstract: While deep reinforcement learning (RL) agents outperform humans on an increasing number of tasks, training them requires data equivalent to decades of human gameplay. Recent hierarchical RL methods have increased sample efficiency by incorporating information inherent to the structure of the decision problem but at the cost of having to discover or use human-annotated sub-goals that guide the lear… ▽ More While deep reinforcement learning (RL) agents outperform humans on an increasing number of tasks, training them requires data equivalent to decades of human gameplay. Recent hierarchical RL methods have increased sample efficiency by incorporating information inherent to the structure of the decision problem but at the cost of having to discover or use human-annotated sub-goals that guide the learning process. We show that intentions of human players, i.e. the precursor of goal-oriented decisions, can be robustly predicted from eye gaze even for the long-horizon sparse rewards task of Montezuma's Revenge - one of the most challenging RL tasks in the Atari2600 game suite. We propose Int-HRL: Hierarchical RL with intention-based sub-goals that are inferred from human eye gaze. Our novel sub-goal extraction pipeline is fully automatic and replaces the need for manual sub-goal annotation by human experts. Our evaluations show that replacing hand-crafted sub-goals with automatically extracted intentions leads to a HRL agent that is significantly more sample efficient than previous methods. △ Less

Submitted 20 June, 2023; originally announced June 2023.

TangleSim: An Agent-based, Modular Simulator for DAG-based Distributed Ledger Technologies

Authors: Bing-Yang Lin, Daria Dziubałtowska, Piotr Macek, Andreas Penzkofer, Sebastian Müller

Abstract: DAG-based DLTs allow for parallel, asynchronous writing access to a ledger. Consequently, the perception of the most recent blocks may differ considerably between nodes, and the underlying network properties of the P2P layer have a direct impact on the performance of the protocol. Moreover, the stronger inter-dependencies of several core components demand a more complex and complete approach to st… ▽ More DAG-based DLTs allow for parallel, asynchronous writing access to a ledger. Consequently, the perception of the most recent blocks may differ considerably between nodes, and the underlying network properties of the P2P layer have a direct impact on the performance of the protocol. Moreover, the stronger inter-dependencies of several core components demand a more complex and complete approach to studying such DLTs. This paper presents an agent-based, open-sourced simulator for large-scale networks that implement the leaderless Tangle 2.0 consensus protocol. Its scope includes modelling the underlying peer-to-peer communication with network topology, package loss, heterogeneous latency, the gossip protocol with reliable broadcast qualities, the underlying DAG-based data structure, and the consensus protocol. The simulator allows us to explore the performance of the protocol in different network environments, as well as different attack scenarios. △ Less

Submitted 2 May, 2023; originally announced May 2023.

Comments: IEEE ICBC 2023, short paper

Mitigation of liveness attacks in DAG-based ledgers

Authors: Darcy Camargo, Andreas Penzkofer, Sebastian Müller, William Sanders

Abstract: The robust construction of the ledger data structure is an essential ingredient for the safe operation of a distributed ledger. While in traditional linear blockchain systems, permission to append to the structure is leader-based, in Directed Acyclic Graph-based ledgers, the writing access can be organised leaderless. However, this leaderless approach relies on fair treatment of non-referenced blo… ▽ More The robust construction of the ledger data structure is an essential ingredient for the safe operation of a distributed ledger. While in traditional linear blockchain systems, permission to append to the structure is leader-based, in Directed Acyclic Graph-based ledgers, the writing access can be organised leaderless. However, this leaderless approach relies on fair treatment of non-referenced blocks, i.e. tips, by honest block issuers. We study the impact of a deviation from the standard tip selection by a subset of block issuers with the aim of halting the confirmation of honest blocks entirely. We provide models on this so-called orphanage of blocks and validate these through open-sourced simulation studies. A critical threshold for the adversary issuance rate is shown to exist, above which the tip pool becomes unstable, while for values below the orphanage decrease exponentially. We study the robustness of the protocol with an expiration time on tips, also called garbage collection, and modification of the parent references per block. △ Less

Submitted 2 May, 2023; originally announced May 2023.

Comments: IEEE ICBC 2023

Report on the energy consumption of the IOTA 2.0 prototype network (GoShimmer 0.8.3) under different testing scenarios

Authors: Louis Helmer, Andreas Penzkofer

Abstract: The high energy consumption of proof of work-based distributed ledgers has become an important environmental concern. Bitcoin, for example, consumes as much energy in a year as a developed country. Alternative consensus mechanisms, such as proof of stake, have been shown to use drastically less energy than proof of work-based DLTs. For example, the IOTA DLT, built upon a directed acyclic graph (DA… ▽ More The high energy consumption of proof of work-based distributed ledgers has become an important environmental concern. Bitcoin, for example, consumes as much energy in a year as a developed country. Alternative consensus mechanisms, such as proof of stake, have been shown to use drastically less energy than proof of work-based DLTs. For example, the IOTA DLT, built upon a directed acyclic graph (DAG) architecture, uses an alternative consensus mechanism that requires significantly less energy than other DLTs. Because the (DLT) space is constantly and rapidly evolving, the question of how much energy DLTs actually consume demands to be continuously studied and answered. Previous research into the energy consumption of the IOTA network has shown that an optimization in the overall protocol correlates to an optimization in energy consumption. The planned IOTA 2.0 update, built upon the GoShimmer research prototype, promises to further optimize the protocol by removing the network's centralized Coordinator. This report presents the results of measuring the energy consumption of a private GoShimmer network while comparing these findings to previous research into the current mainnet, which is called Chrysalis. The main findings of this report are that the IOTA 2.0 research prototype shows both improvements and increase in the energy consumption metrics compared to the Chrysalis network. Additionally, this report defines a model to estimate the total annual energy consumption of an IOTA network. This model should be significant for future research as it enables a way to estimate the total cost of running the IOTA network as well as its carbon emissions. Moreover, having an annual power consumption metric allows for better objective comparisons to different DLTs. △ Less

Submitted 25 October, 2022; originally announced October 2022.

Robustness of the Tangle 2.0 Consensus

Authors: Bing-Yang Lin, Daria Dziubałtowska, Piotr Macek, Andreas Penzkofer, Sebastian Müller

Abstract: In this paper, we investigate the performance of the Tangle 2.0 consensus protocol in a Byzantine environment. We use an agent-based simulation model that incorporates the main features of the Tangle 2.0 consensus protocol. Our experimental results demonstrate that the Tangle 2.0 protocol is robust to the bait-and-switch attack up to the theoretical upper bound of the adversary's 33% voting weight… ▽ More In this paper, we investigate the performance of the Tangle 2.0 consensus protocol in a Byzantine environment. We use an agent-based simulation model that incorporates the main features of the Tangle 2.0 consensus protocol. Our experimental results demonstrate that the Tangle 2.0 protocol is robust to the bait-and-switch attack up to the theoretical upper bound of the adversary's 33% voting weight. We further show that the common coin mechanism in Tangle 2.0 is necessary for robustness against powerful adversaries. Moreover, the experimental results confirm that the protocol can achieve around 1s confirmation time in typical scenarios and that the confirmation times of non-conflicting transactions are not affected by the presence of conflicts. △ Less

Submitted 24 August, 2022; v1 submitted 17 August, 2022; originally announced August 2022.

Tangle 2.0 Leaderless Nakamoto Consensus on the Heaviest DAG

Authors: Sebastian Müller, Andreas Penzkofer, Nikita Polyanskii, Jonas Theis, William Sanders, Hans Moog

Abstract: We introduce the theoretical foundations of the Tangle 2.0, a probabilistic leaderless consensus protocol based on a directed acyclic graph (DAG) called the Tangle. The Tangle naturally succeeds the blockchain as its next evolutionary step as it offers features suited to establish more efficient and scalable distributed ledger solutions. Consensus is no longer found in the longest chain but on t… ▽ More We introduce the theoretical foundations of the Tangle 2.0, a probabilistic leaderless consensus protocol based on a directed acyclic graph (DAG) called the Tangle. The Tangle naturally succeeds the blockchain as its next evolutionary step as it offers features suited to establish more efficient and scalable distributed ledger solutions. Consensus is no longer found in the longest chain but on the heaviest DAG, where PoW is replaced by a stake- or reputation-based weight function. The DAG structure and the underlying Reality-based UTXO Ledger allow parallel validation of transactions without the need for total ordering. Moreover, it enables the removal of the intermediary of miners and validators, allowing a pure two-step process that follows the \emph{propose-vote} paradigm at the node level and not at the validator level. We propose a framework to analyse liveness and safety under different communication and adversary models. This allows providing impossibility results in some edge cases and in the asynchronous communication model. We provide formal proof of the security of the protocol assuming a common random coin. △ Less

Submitted 12 October, 2022; v1 submitted 4 May, 2022; originally announced May 2022.

Comments: revised version, to appear in IEEE Access

Reality-based UTXO Ledger

Authors: Sebastian Müller, Andreas Penzkofer, Nikita Polyanskii, Jonas Theis, William Sanders, Hans Moog

Abstract: The Unspent Transaction Output (UTXO) model is commonly used in the field of Distributed Ledger Technology (DLT) to transfer value between participants. One of its advantages is that it allows parallel processing of transactions, as independent transactions can be added in any order. This property of order invariance and parallelisability has potential benefits in terms of scalability. However, si… ▽ More The Unspent Transaction Output (UTXO) model is commonly used in the field of Distributed Ledger Technology (DLT) to transfer value between participants. One of its advantages is that it allows parallel processing of transactions, as independent transactions can be added in any order. This property of order invariance and parallelisability has potential benefits in terms of scalability. However, since the UTXO Ledger is an append-only data structure, this advantage is compromised through the presence of conflicting transactions. We propose an extended UTXO Ledger model that optimistically updates the ledger and keeps track of the dependencies of the possible conflicts. In the presence of a conflict resolution mechanism, we propose a method to reduce the extended ledger back to a consistent UTXO Ledger. △ Less

Submitted 7 August, 2023; v1 submitted 3 May, 2022; originally announced May 2022.

Comments: revised version, accepted in ACM DLT

Salt-based autopeering for DLT-networks

Authors: Sebastian Müller, Angelo Capossele, Bartosz Kuśmierz, Vivian Lin, Hans Moog, Andreas Penzkofer, Olivia Saa, William Sanders, Wolfgang Welz

Abstract: The security of any Distributed Ledger Technology (DLT) depends on the safety of the network layer. Much effort has been put into understanding the consensus layer of DLTs. However, many network layer designs seem ad-hoc and lack a careful analysis of the influence of the design decisions on the whole DLT system. We propose a salt-based automated neighbor selection protocol that shows the inherent… ▽ More The security of any Distributed Ledger Technology (DLT) depends on the safety of the network layer. Much effort has been put into understanding the consensus layer of DLTs. However, many network layer designs seem ad-hoc and lack a careful analysis of the influence of the design decisions on the whole DLT system. We propose a salt-based automated neighbor selection protocol that shows the inherent tradeoffs of certain design decisions and allows a quantitative treatment of some network topology requirements. This example may serve as a design framework and facilitate future research. We provide a selection of results from simulations to highlight some tradeoffs in the design decisions. △ Less

Submitted 3 November, 2021; originally announced November 2021.

Comments: 4 pages

Journal ref: 2021 3rd Conference on Blockchain Research Applications for Innovative Networks and Services (BRAINS)

Impact of delay classes on the data structure in IOTA

Authors: Andreas Penzkofer, Olivia Saa, Daria Dziubałtowska

Abstract: In distributed ledger technologies (DLTs) with a directed acyclic graph (DAG) data structure, a message-issuing node can decide where to append that message and, consequently, how to grow the DAG. This DAG data structure can typically be decomposed into two pools of messages: referenced messages and unreferenced messages (tips). The selection of the parent messages to which a node appends the mess… ▽ More In distributed ledger technologies (DLTs) with a directed acyclic graph (DAG) data structure, a message-issuing node can decide where to append that message and, consequently, how to grow the DAG. This DAG data structure can typically be decomposed into two pools of messages: referenced messages and unreferenced messages (tips). The selection of the parent messages to which a node appends the messages it issues, depends on which messages it considers as tips. However, the exact time that a message enters the tip pool of a node depends on the delay of that message. In previous works, it was considered that messages have the same or similar delay; however, this generally may not be the case. We introduce the concept of classes of delays, where messages belonging to a certain class have a specific delay, and where these classes coexist in the DAG. We provide a general model that predicts the tip pool size for any finite number of different classes. This categorisation and model is applied to the first iteration of the IOTA 2.0 protocol (a.k.a. Coordicide), where two distinct classes, namely value and data messages, coexist. We show that the tip pool size depends strongly on the dominating class that is present. Finally, we provide a methodology for controlling the tip pool size by dynamically adjusting the number of references a message creates. △ Less

Submitted 12 October, 2021; originally announced October 2021.

Comments: Presented at CBT2021 - 5th Cryptocurrencies and Blockchain Technology workshop https://www.esorics2021.org/program/

On Fairness in Voting Consensus Protocols

Authors: Sebastian Müller, Andreas Penzkofer, Darcy Camargo, Olivia Saa

Abstract: Voting algorithms have been widely used as consensus protocols in the realization of fault-tolerant systems. These algorithms are best suited for distributed systems of nodes with low computational power or heterogeneous networks, where different nodes may have different levels of reputation or weight. Our main contribution is the construction of a fair voting protocol in the sense that the influe… ▽ More Voting algorithms have been widely used as consensus protocols in the realization of fault-tolerant systems. These algorithms are best suited for distributed systems of nodes with low computational power or heterogeneous networks, where different nodes may have different levels of reputation or weight. Our main contribution is the construction of a fair voting protocol in the sense that the influence of the eventual outcome of a given participant is linear in its weight. Specifically, the fairness property guarantees that any node can actively participate in the consensus finding even with low resources or weight. We investigate effects that may arise from weighted voting, such as loss of anonymity, centralization, scalability, and discuss their relevance to protocol design and implementation. △ Less

Submitted 15 December, 2020; originally announced December 2020.

Comments: conference paper, SAI Computing21. arXiv admin note: text overlap with arXiv:2006.00928

Fast Probabilistic Consensus with Weighted Votes

Authors: Sebastian Müller, Andreas Penzkofer, Bartosz Kuśmierz, Darcy Camargo, William J. Buchanan

Abstract: The fast probabilistic consensus (FPC) is a voting consensus protocol that is robust and efficient in Byzantine infrastructure. We propose an adaption of the FPC to a setting where the voting power is proportional to the nodes reputations. We model the reputation using a Zipf law and show using simulations that the performance of the protocol in Byzantine infrastructure increases with the Zipf exp… ▽ More The fast probabilistic consensus (FPC) is a voting consensus protocol that is robust and efficient in Byzantine infrastructure. We propose an adaption of the FPC to a setting where the voting power is proportional to the nodes reputations. We model the reputation using a Zipf law and show using simulations that the performance of the protocol in Byzantine infrastructure increases with the Zipf exponent. Moreover, we propose several improvements of the FPC that decrease the failure rates significantly and allow the protocol to withstand adversaries with higher weight. We distinguish between cautious and berserk strategies of the adversaries and propose an efficient method to detect the more harmful berserk strategies. Our study refers at several points to a specific implementation of the IOTA protocol, but the principal results hold for general implementations of reputation models. △ Less

Submitted 1 June, 2020; originally announced June 2020.

Report number: Conference: FTC 2020 - Future Technologies Conference 2020

Parasite Chain Detection in the IOTA Protocol

Authors: Andreas Penzkofer, Bartosz Kusmierz, Angelo Capossele, William Sanders, Olivia Saa

Abstract: In recent years several distributed ledger technologies based on directed acyclic graphs (DAGs) have appeared on the market. Similar to blockchain technologies, DAG-based systems aim to build an immutable ledger and are faced with security concerns regarding the irreversibility of the ledger state. However, due to their more complex nature and recent popularity, the study of adversarial actions ha… ▽ More In recent years several distributed ledger technologies based on directed acyclic graphs (DAGs) have appeared on the market. Similar to blockchain technologies, DAG-based systems aim to build an immutable ledger and are faced with security concerns regarding the irreversibility of the ledger state. However, due to their more complex nature and recent popularity, the study of adversarial actions has received little attention so far. In this paper we are concerned with a particular type of attack on the IOTA cryptocurrency, more specifically a Parasite Chain attack that attempts to revert the history stored in the DAG structure, also called the Tangle. In order to improve the security of the Tangle, we present a detection mechanism for this type of attack. In this mechanism, we embrace the complexity of the DAG structure by sampling certain aspects of it, more particularly the distribution of the number of approvers. We initially describe models that predict the distribution that should be expected for a Tangle without any malicious actors. We then introduce metrics that compare this reference distribution with the measured distribution. Upon detection, measures can then be taken to render the attack unsuccessful. We show that due to a form of the Parasite Chain that is different from the main Tangle it is possible to detect certain types of malicious chains. We also show that although the attacker may change the structure of the Parasite Chain to avoid detection, this is done so at a significant cost since the attack is rendered less efficient. △ Less

Submitted 28 April, 2020; originally announced April 2020.

Comments: Accepted for publication at Tokenomics 2020, Toulouse, France

Properties of the Tangle for Uniform Random and Random Walk Tip Selection

Authors: Bartosz Kusmierz, William Sanders, Andreas Penzkofer, Angelo Capossele, Alon Gal

Abstract: The growing number of applications for distributed ledger technologies is driving both industry and academia to solve the limitations of blockchain, particularly its scalability issues. Recent distributed ledger technologies have replaced the blockchain linear structure with a more flexible directed acyclic graph in an attempt to accommodate a higher throughput. Despite the fast-growing diffusion… ▽ More The growing number of applications for distributed ledger technologies is driving both industry and academia to solve the limitations of blockchain, particularly its scalability issues. Recent distributed ledger technologies have replaced the blockchain linear structure with a more flexible directed acyclic graph in an attempt to accommodate a higher throughput. Despite the fast-growing diffusion of directed acyclic graph based distributed ledger technologies, researchers lack a basic understanding of their behavior. In this paper we analyze the Tangle, a directed acyclic graph that is used (with certain modifications) in various protocols such as IOTA, Byteball, Avalanche or SPECTRE. Our contribution is threefold. First, we run simulations in a continuous-time model to examine tip count stability and cumulative weight evolution while varying the rate of incoming transactions. In particular we confirm analytical predictions on the number of tips with uniform random tip selection strategy. Second, we show how different tip selection algorithms affect the growth of the Tangle. Moreover, we explain these differences by analyzing the spread of exit probabilities of random walks. Our findings confirm analytically derived predictions and provide novel insights on the different phases of growth of cumulative weight as well as on the average time difference for a transaction to receive its first approval when using distinct tip selection algorithms. Lastly, we analyze simulation overhead and performance as a function of Tangle size and compare results for different tip selection algorithms. △ Less

Submitted 21 January, 2020; originally announced January 2020.

Comments: Published in: 2019 IEEE International Conference on Blockchain (Blockchain)

Journal ref: 2019 IEEE International Conference on Blockchain (Blockchain)

Robustness and efficiency of leaderless probabilistic consensus protocols within Byzantine infrastructures

Authors: Angelo Capossele, Sebastian Mueller, Andreas Penzkofer

Abstract: This paper investigates leaderless binary majority consensus protocols with low computational complexity in noisy Byzantine infrastructures. Using computer simulations, we show that explicit randomization of the consensus protocol can significantly increase the robustness towards faulty and malicious nodes. We identify the optimal amount of randomness for various Byzantine attack strategies on dif… ▽ More This paper investigates leaderless binary majority consensus protocols with low computational complexity in noisy Byzantine infrastructures. Using computer simulations, we show that explicit randomization of the consensus protocol can significantly increase the robustness towards faulty and malicious nodes. We identify the optimal amount of randomness for various Byzantine attack strategies on different kinds of network topologies. △ Less

Submitted 20 November, 2019; originally announced November 2019.

Comments: 21 pages, 20 figures

ACM Class: C.2.4; C.4