David Aparício
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
David Aparício has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Pedro Ribeiro
CRACS-INESC TEC, Faculdade de Ciências, Universidade do Porto, Porto, Portugal
12 shared publications - Fernando Silva
CRACS-INESC TEC, Faculdade de Ciências, Universidade do Porto, Porto, Portugal
9 shared publications Fernando Silva
2 shared publications- Jacopo Bono
Feedzai, Portugal
2 shared publications Jorge Silva
University of Porto, Porto, Portugal
2 shared publications- Jorge Silva
CRACS/INESC TEC & Department of Computer Science, Faculty of Sciences, University of Porto (DCC-FCUP), Porto, Portugal
2 shared publications
Latest Publications
- Open Access
- 25 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 133-140
The GANfather: Controllable generation of malicious activity to improve defence systems
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.
- 25 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 133-140
Open AccessThe GANfather: Controllable generation of malicious activity to improve defence systems
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.
- Open Access
- 25 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 176-184
From random-walks to graph-sprints: a low-latency node embedding framework on continuous-time dynamic graphs
- 25 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 176-184
Open AccessFrom random-walks to graph-sprints: a low-latency node embedding framework on continuous-time dynamic graphs
- Open Access
Anti-Money Laundering Alert Optimization Using Machine Learning with Graphs
Money laundering is a global problem that concerns legitimizing proceeds from serious felonies (1.7-4 trillion euros annually) such as drug dealing, human trafficking, or corruption. The anti-money laundering systems deployed by financial institutions typically comprise rules aligned with regulatory frameworks. Human investigators review the alerts and report suspicious cases. Such systems suffer from high false-positive rates, undermining their effectiveness and resulting in high operational costs. We propose a machine learning triage model, which complements the rule-based system and learns to predict the risk of an alert accurately. Our model uses both entity-centric engineered features and attributes characterizing inter-entity relations in the form of graph-based features. We leverage time windows to construct the dynamic graph, optimizing for time and space efficiency. We validate our model on a real-world banking dataset and show how the triage model can reduce the number of false positives by 80% while detecting over 90% of true positives. In this way, our model can significantly improve anti-money laundering operations.
Open AccessAnti-Money Laundering Alert Optimization Using Machine Learning with Graphs
Money laundering is a global problem that concerns legitimizing proceeds from serious felonies (1.7-4 trillion euros annually) such as drug dealing, human trafficking, or corruption. The anti-money laundering systems deployed by financial institutions typically comprise rules aligned with regulatory frameworks. Human investigators review the alerts and report suspicious cases. Such systems suffer from high false-positive rates, undermining their effectiveness and resulting in high operational costs. We propose a machine learning triage model, which complements the rule-based system and learns to predict the risk of an alert accurately. Our model uses both entity-centric engineered features and attributes characterizing inter-entity relations in the form of graph-based features. We leverage time windows to construct the dynamic graph, optimizing for time and space efficiency. We validate our model on a real-world banking dataset and show how the triage model can reduce the number of false positives by 80% while detecting over 90% of true positives. In this way, our model can significantly improve anti-money laundering operations.
- Open Access
- 30 March 2020
- Published by Association for Computing Machinery (ACM)
- p. 1852-1860
FOCAS
Scientific impact is commonly associated with the number of citations received. However, an author can easily boost his own citation count by (i) publishing articles that cite his own previous work (self-citations), (ii) having co-authors citing his work (co-author citations), or (iii) exchanging citations with authors from other research groups (reciprocated citations). Even though these friendly citations inflate an author's perceived scientific impact, author ranking algorithms do not normally address them. They, at most, remove self-citations. Here we present Friends-Only Citations AnalySer (FOCAS), a method that identifies friendly citations and reduces their negative effect in author ranking algorithms. FOCAS combines the author citation network with the co-authorship network in order to measure author proximity and penalises citations between friendly authors. FOCAS is general and can be regarded as an independent module applied while running (any) PageRank-like author ranking algorithm. FOCAS can be tuned to use three different criteria, namely authors' distance, citation frequency, and citation recency, or combinations of these. We evaluate and compare FOCAS against eight state-of-the-art author ranking algorithms. We compare their rankings with a ground-truth of best paper awards. We test our hypothesis on a citation and co-authorship network comprised of seven Information Retrieval top-conferences. We observed that FOCAS improved author rankings by 25% on average and, in one case, leads to a gain of 46%.
- 30 March 2020
- Published by Association for Computing Machinery (ACM)
- p. 1852-1860
Open AccessFOCAS
Scientific impact is commonly associated with the number of citations received. However, an author can easily boost his own citation count by (i) publishing articles that cite his own previous work (self-citations), (ii) having co-authors citing his work (co-author citations), or (iii) exchanging citations with authors from other research groups (reciprocated citations). Even though these friendly citations inflate an author's perceived scientific impact, author ranking algorithms do not normally address them. They, at most, remove self-citations. Here we present Friends-Only Citations AnalySer (FOCAS), a method that identifies friendly citations and reduces their negative effect in author ranking algorithms. FOCAS combines the author citation network with the co-authorship network in order to measure author proximity and penalises citations between friendly authors. FOCAS is general and can be regarded as an independent module applied while running (any) PageRank-like author ranking algorithm. FOCAS can be tuned to use three different criteria, namely authors' distance, citation frequency, and citation recency, or combinations of these. We evaluate and compare FOCAS against eight state-of-the-art author ranking algorithms. We compare their rankings with a ground-truth of best paper awards. We test our hypothesis on a citation and co-authorship network comprised of seven Information Retrieval top-conferences. We observed that FOCAS improved author rankings by 25% on average and, in one case, leads to a gain of 46%.
- Open Access
- 23 September 2019
- Published by Springer Nature in Applied Network Science
- Vol. 4 (1) , 1-15
Feature-enriched author ranking in incomplete networks
Evaluating scientists based on their scientific production is a controversial topic. Nevertheless, bibliometrics and algorithmic approaches can assist traditional peer review in numerous tasks, such as attributing research grants, deciding scientific committees, or choosing faculty promotions. Traditional bibliometrics rank individual entities (e.g., researchers, journals, faculties) without looking at the whole data (i.e., the whole network). Network algorithms, such as PageRank, have been used to measure node importance in a network, and have been applied to author ranking. However, traditional PageRank only uses network topology and ignores relevant features of scientific collaborations. Multiple extensions of PageRank have been proposed, more suited for author ranking. These methods enrich the network with information about the author’s productivity or the venue and year of the publication/citation. Most state-of-the-art (STOA) feature-enriched methods either ignore or do not combine effectively this information. Furthermore, STOA algorithms typically disregard that the full network is not known for most real-world cases.Here we describe OTARIOS, an author ranking method recently developed by us, which combines multiple publication/citation criteria (i.e., features) to evaluate authors. OTARIOS divides the original network into two subnetworks,insidersandoutsiders, which is an adequate representation of citation networks with missing information. We evaluate OTARIOS on a set of five real networks, each with publications in distinct areas of Computer Science, and compare it against STOA methods. When matching OTARIOS’ produced ranking with a ground-truth ranking (comprised of best paper award nominations), we observe that OTARIOS is >30%more accurate than traditional PageRank (i.e., topology based method) and >20%more accurate than STOA (i.e., competing feature enriched methods). We obtain the best results when OTARIOS considers (i) the author’s publication volume and publication recency, (ii) how recently the author’s work is being cited by outsiders, and (iii) how recently the author’s work is being cited by insiders and how individual he is. Our results showcase (a) the importance of efficiently combining relevant features and (b) how to adequately perform author ranking in incomplete networks.
- 23 September 2019
- Published by Springer Nature in Applied Network Science
- Vol. 4 (1) , 1-15
Open AccessFeature-enriched author ranking in incomplete networks
Evaluating scientists based on their scientific production is a controversial topic. Nevertheless, bibliometrics and algorithmic approaches can assist traditional peer review in numerous tasks, such as attributing research grants, deciding scientific committees, or choosing faculty promotions. Traditional bibliometrics rank individual entities (e.g., researchers, journals, faculties) without looking at the whole data (i.e., the whole network). Network algorithms, such as PageRank, have been used to measure node importance in a network, and have been applied to author ranking. However, traditional PageRank only uses network topology and ignores relevant features of scientific collaborations. Multiple extensions of PageRank have been proposed, more suited for author ranking. These methods enrich the network with information about the author’s productivity or the venue and year of the publication/citation. Most state-of-the-art (STOA) feature-enriched methods either ignore or do not combine effectively this information. Furthermore, STOA algorithms typically disregard that the full network is not known for most real-world cases.Here we describe OTARIOS, an author ranking method recently developed by us, which combines multiple publication/citation criteria (i.e., features) to evaluate authors. OTARIOS divides the original network into two subnetworks,insidersandoutsiders, which is an adequate representation of citation networks with missing information. We evaluate OTARIOS on a set of five real networks, each with publications in distinct areas of Computer Science, and compare it against STOA methods. When matching OTARIOS’ produced ranking with a ground-truth ranking (comprised of best paper award nominations), we observe that OTARIOS is >30%more accurate than traditional PageRank (i.e., topology based method) and >20%more accurate than STOA (i.e., competing feature enriched methods). We obtain the best results when OTARIOS considers (i) the author’s publication volume and publication recency, (ii) how recently the author’s work is being cited by outsiders, and (iii) how recently the author’s work is being cited by insiders and how individual he is. Our results showcase (a) the importance of efficiently combining relevant features and (b) how to adequately perform author ranking in incomplete networks.
- 13 February 2019
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 35 (18) , 3527-3529
Temporal network alignment via GoT-WAVE
Network alignment (NA) finds conserved regions between two networks. NA methods optimize node conservation (NC) and edge conservation. Dynamic graphlet degree vectors are a state-of-the-art dynamic NC measure, used within the fastest and most accurate NA method for temporal networks: DynaWAVE. Here, we use graphlet-orbit transitions (GoTs), a different graphlet-based measure of temporal node similarity, as a new dynamic NC measure within DynaWAVE, resulting in GoT-WAVE. On synthetic networks, GoT-WAVE improves DynaWAVE’s accuracy by 30% and speed by 64%. On real networks, when optimizing only dynamic NC, the methods are complementary. Furthermore, only GoT-WAVE supports directed edges. Hence, GoT-WAVE is a promising new temporal NA algorithm, which efficiently optimizes dynamic NC. We provide a user-friendly user interface and source code for GoT-WAVE. http://www.dcc.fc.up.pt/got-wave/ Supplementary data are available at Bioinformatics online.
- 13 February 2019
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 35 (18) , 3527-3529
Temporal network alignment via GoT-WAVE
Network alignment (NA) finds conserved regions between two networks. NA methods optimize node conservation (NC) and edge conservation. Dynamic graphlet degree vectors are a state-of-the-art dynamic NC measure, used within the fastest and most accurate NA method for temporal networks: DynaWAVE. Here, we use graphlet-orbit transitions (GoTs), a different graphlet-based measure of temporal node similarity, as a new dynamic NC measure within DynaWAVE, resulting in GoT-WAVE. On synthetic networks, GoT-WAVE improves DynaWAVE’s accuracy by 30% and speed by 64%. On real networks, when optimizing only dynamic NC, the methods are complementary. Furthermore, only GoT-WAVE supports directed edges. Hence, GoT-WAVE is a promising new temporal NA algorithm, which efficiently optimizes dynamic NC. We provide a user-friendly user interface and source code for GoT-WAVE. http://www.dcc.fc.up.pt/got-wave/ Supplementary data are available at Bioinformatics online.
- 2 December 2018
- Published by Springer Nature
- p. 143-154
OTARIOS: OpTimizing Author Ranking with Insiders/Outsiders Subnetworks
- 2 December 2018
- Published by Springer Nature
- p. 143-154
OTARIOS: OpTimizing Author Ranking with Insiders/Outsiders Subnetworks
- Open Access
- 18 October 2018
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 13 (10) , e0205497
Graphlet-orbit Transitions (GoT): A fingerprint for temporal network comparison
Given a set of temporal networks, from different domains and with different sizes, how can we compare them? Can we identify evolutionary patterns that are both (i) characteristic and (ii) meaningful? We address these challenges by introducing a novel temporal and topological network fingerprint named Graphlet-orbit Transitions (GoT). We demonstrate that GoT provides very rich and interpretable network characterizations. Our work puts forward an extension of graphlets and uses the notion of orbits to encapsulate the roles of nodes in each subgraph. We build a transition matrix that keeps track of the temporal trajectory of nodes in terms of their orbits, therefore describing their evolution. We also introduce a metric (OTA) to compare two networks when considering these matrices. Our experiments show that networks representing similar systems have characteristic orbit transitions. GoT correctly groups synthetic networks pertaining to well-known graph models more accurately than competing static and dynamic state-of-the-art approaches by over 30%. Furthermore, our tests on real-world networks show that GoT produces highly interpretable results, which we use to provide insight into characteristic orbit transitions.
- 18 October 2018
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 13 (10) , e0205497
Open AccessGraphlet-orbit Transitions (GoT): A fingerprint for temporal network comparison
Given a set of temporal networks, from different domains and with different sizes, how can we compare them? Can we identify evolutionary patterns that are both (i) characteristic and (ii) meaningful? We address these challenges by introducing a novel temporal and topological network fingerprint named Graphlet-orbit Transitions (GoT). We demonstrate that GoT provides very rich and interpretable network characterizations. Our work puts forward an extension of graphlets and uses the notion of orbits to encapsulate the roles of nodes in each subgraph. We build a transition matrix that keeps track of the temporal trajectory of nodes in terms of their orbits, therefore describing their evolution. We also introduce a metric (OTA) to compare two networks when considering these matrices. Our experiments show that networks representing similar systems have characteristic orbit transitions. GoT correctly groups synthetic networks pertaining to well-known graph models more accurately than competing static and dynamic state-of-the-art approaches by over 30%. Furthermore, our tests on real-world networks show that GoT produces highly interpretable results, which we use to provide insight into characteristic orbit transitions.
- Open Access
- 8 September 2018
- Published by Springer Nature
Graphlet-orbit Transitions v1
Given a set of temporal networks, from different domains and with different sizes, how can we compare them? Can we identify evolutionary patterns that are both (i) characteristic and (ii) meaningful? We address these challenges by introducing a novel temporal and topological network fingerprint named Graphlet-orbit Transitions (GoT). We demonstrate that GoT provides very rich and interpretable network characterizations. Our work puts forward an extension of graphlets and uses the notion of orbits to encapsulate the roles of nodes in each subgraph. We build a transition matrix that keeps track of the temporal trajectory of nodes in terms of their orbits, therefore describing their evolution. We also introduce a metric (OTA) to compare two networks when considering these matrices. Our experiments show that networks representing similar systems have characteristic orbit transitions. GoT correctly groups synthetic networks pertaining to well-known graph models more accurately than competing static and dynamic state-of-the-art approaches by over 30%. Furthermore, our tests on real-world networks show that GoT produces highly interpretable results, which we use to provide insight into characteristic orbit transitions.
- 8 September 2018
- Published by Springer Nature
Open AccessGraphlet-orbit Transitions v1
Given a set of temporal networks, from different domains and with different sizes, how can we compare them? Can we identify evolutionary patterns that are both (i) characteristic and (ii) meaningful? We address these challenges by introducing a novel temporal and topological network fingerprint named Graphlet-orbit Transitions (GoT). We demonstrate that GoT provides very rich and interpretable network characterizations. Our work puts forward an extension of graphlets and uses the notion of orbits to encapsulate the roles of nodes in each subgraph. We build a transition matrix that keeps track of the temporal trajectory of nodes in terms of their orbits, therefore describing their evolution. We also introduce a metric (OTA) to compare two networks when considering these matrices. Our experiments show that networks representing similar systems have characteristic orbit transitions. GoT correctly groups synthetic networks pertaining to well-known graph models more accurately than competing static and dynamic state-of-the-art approaches by over 30%. Furthermore, our tests on real-world networks show that GoT produces highly interpretable results, which we use to provide insight into characteristic orbit transitions.