Full publication list

Currently viewing only publications on theme “Evolution”.
Click here to reset the filter
Currently viewing only publications on theme “Privacy”.
Click here to reset the filter

PhD Thesis

Machine learning for autonomous privacy preservation in participatory sensing and smart cities.
Stefano Bennati. 2018

Journal contributions

Modelling imperfect knowledge via location semantics for realistic privacy risks estimation in trajectory data.
Stefano Bennati, Aleksandra Kovacevic. 2022

Published in: Scientific Reports

Keywords: Anonymization, Location Data, Privacy Risk Estimation, Location Semantics, Equivalence Areas, Imperfect Knowledge.

In this paper, titled “Modelling imperfect knowledge via location semantics for realistic privacy risks estimation in trajectory data” we propose a privacy risk estimation framework that builds on the semantic context associated to location data. By moving away from the assumption that adversaries have perfect knowledge, we can obtain realistic risk estimates and optimize the trade-off between privacy and utility.

How learning can change the course of evolution.
Stefano Bennati, Leonel Aguilar, Dirk Helbing. 2019

Published in: PlosOne


Keywords: Baldwin effect, Foraging, Specialist, Generalist, Environmental variability, Evolution, Agent-based modeling, Learning, Neural network, Reinforcement learning


The interaction between phenotypic plasticity, e.g. learning, and evolution is an important topic both in Evolutionary Biology and Machine Learning. The evolution of learning is commonly studied in Evolutionary Biology, while the use of an evolutionary process to improve learning is of interest to the field of Machine Learning. This paper takes a different point of view by studying the effect of learning on the evolutionary process, the so-called Baldwin effect. A well-studied result in the literature about the Baldwin effect is that learning affects the speed of convergence of the evolutionary process towards some genetic configuration, which corresponds to the environment-induced plastic response. This paper demonstrates that learning can change the outcome of evolution, i.e., lead to a genetic configuration that does not correspond to the plastic response. Results are obtained both analytically and experimentally by means of an agent-based model of a foraging task, in an environment where the distribution of resources follows seasonal cycles and the foraging success on different resource types is conditioned by trade-offs that can be evolved and learned. This paper attempts to answer a question that has been overlooked: whether learning has an effect on what genotypic traits are evolved, i.e. the selection of a trait that enables a plastic response changes the selection pressure on a different trait, in what could be described as co-evolution between different traits in the same genome.
How learning can change the course of evolution.
Stefano Bennati, Leonel Aguilar, Dirk Helbing. 2019

Published in: PlosOne


Keywords: Baldwin effect, Foraging, Specialist, Generalist, Environmental variability, Evolution, Agent-based modeling, Learning, Neural network, Reinforcement learning


The interaction between phenotypic plasticity, e.g. learning, and evolution is an important topic both in Evolutionary Biology and Machine Learning. The evolution of learning is commonly studied in Evolutionary Biology, while the use of an evolutionary process to improve learning is of interest to the field of Machine Learning. This paper takes a different point of view by studying the effect of learning on the evolutionary process, the so-called Baldwin effect. A well-studied result in the literature about the Baldwin effect is that learning affects the speed of convergence of the evolutionary process towards some genetic configuration, which corresponds to the environment-induced plastic response. This paper demonstrates that learning can change the outcome of evolution, i.e., lead to a genetic configuration that does not correspond to the plastic response. Results are obtained both analytically and experimentally by means of an agent-based model of a foraging task, in an environment where the distribution of resources follows seasonal cycles and the foraging success on different resource types is conditioned by trade-offs that can be evolved and learned. This paper attempts to answer a question that has been overlooked: whether learning has an effect on what genotypic traits are evolved, i.e. the selection of a trait that enables a plastic response changes the selection pressure on a different trait, in what could be described as co-evolution between different traits in the same genome.
Volunteers in the Smart City: Comparison of Contribution Strategies on Human-Centered Measures.
Stefano Bennati, Ivana Dusparic, Rhythima Shinde, Catholijn Jonker. 2018

 

Published in: Sensors DOI

Keywords: Participatory Sensing , Smart Cities , Public Good , Privacy , Fairness

This paper introduces a scenario-independent design principle for developing smart city algorithms, based on the theory of public goods and voluntary contribution games.
Voluntary contributions are well suited for modeling scenarios where a common resource, i.e. a smart-city service, depends on the users contributing resources, e.g. user-generated data, energy surplus.
The choices of users are modeled by contribution strategies, algorithms that determine which users should contribute to the system at each point in time, based on the state of the system and a set of system requirements.
By acting on the decision of whether to participate, a contribution strategy is independent of the characteristics of the resource, hence it can be combined with existing mechanism that work at the content level, e.g., privacy-preserving algorithms.
A simulation framework is developed that allows a comparison of multiple algorithms for resource contribution on the application scenarios of traffic congestion and charging of electric vehicles, to address the research question: how do different contribution strategies compare in terms of privacy, fairness and social welfare?
Results show that trade-offs between measures and algorithms repeat across the application scenarios, hence this work can be of interest to designers of smart city applications and services that look for a guideline for the choice of algorithm, given specific scenario priorities in terms of different measures.

Privacy-enhancing Aggregation of Internet of Things Data via Sensors Grouping.
Stefano Bennati, Evangelos Pournaras. 2017

Published in: Sustainable Cities and Society DOI


Keywords: privacy, Internet of Things, Smart City, network, sensor, grouping, agent, aggregation


Big data collection practices using Internet of Things (IoT) pervasive technologies are often
privacy-intrusive and result in surveillance, profiling, and discriminatory actions over citizens that in turn undermine the participation of citizens to the development of sustainable smart cities. Nevertheless, real-time data analytics and aggregate information from IoT devices
open up tremendous opportunities for managing and regulating smart city infrastructures in a more efficient and sustainable way. The privacy-enhancing aggregation of distributed sensor data, such as residential energy consumption or traffic information, is the research focus and challenge tackled in this paper. Citizens have the option to choose their privacy level by reducing the quality of the shared data at a cost of a lower accuracy in data analytics services. A baseline scenario is considered in which IoT sensor data are shared directly with an untrustworthy central aggregator. A grouping mechanism is introduced that improves privacy by sharing data aggregated first at a group level compared to a baseline scenario in which each individual shares data directly to the central aggregator. Group-level aggregation obfuscates sensor data of individuals, in a similar fashion as differential privacy and homomorphic encryption schemes, thus inference of privacy-sensitive information from single sensors becomes computationally harder compared to the baseline scenario. The proposed system and its generic applicability are evaluated using real-world data from two smart city pilot projects. Privacy under grouping increases, while preserving the accuracy of the baseline scenario. Intra-group influences of privacy by one group member on the other ones are measured and fairness on privacy is found to be maximized between group members with similar privacy choices. Several grouping strategies are compared. Grouping by proximity of privacy choices provides the highest privacy gains. The implications of the strategy on the design of incentives mechanisms are discussed.
On the Role of Collective Sensing and Evolution in Group Formation.
Stefano Bennati. 2018

Published in: Swarm Intelligence DOI


Keywords: Collective sensing, Natural selection, Foraging, Agent-based modeling


Collective sensing is an emergent phenomenon which enables individuals to estimate a hidden property of the environment through the observation of social interactions. Previous work on collective sensing shows that gregarious individuals obtain an evolutionary advantage by exploiting collective sensing when competing against solitary individuals.
This work addresses the question of whether collective sensing allows for the emergence of groups from a population of individuals without predetermined behaviors. Experiments are run in an agent-based evolutionary model of a foraging task, where the fitness of the agents depends on their foraging strategy. Agents compete for the same limited resources and neither the environment nor inter-group dynamics benefit groups over individuals. The foraging strategy of agents is determined by a model-free neural network, which leaves agent behavior unrestricted.
Experiments demonstrate that gregarious behavior is not the evolutionary- fittest strategy if resources are abundant, thus invalidating previous findings in a specific region of the parameter space. In other words, resource scarcity makes gregarious behavior so valuable as to make up for the increased competition over the few available resources. This result is obtained with a model-free approach which allows evolution to select from an unconstrained set of behavioral models. Furthermore, it is shown that a population of solitary agents can evolve gregarious behavior in response to a sudden scarcity of resources, thus individuating a possible mechanism that leads to gregarious behavior in nature.

Conference contributions

Can Location Data Truly Be Anonymized?
Stefano Bennati, Engin Bozdag. 2023
Presented at the 2023 International Privacy Engineering Workshop (IWPE).
Link to the slides
Incentivized Data Sharing via Group-level Privacy-preservation.
Stefano Bennati, Evangelos Pournaras. 2017
Presented at the 2017 International Conference on Computational Social Science (IC2S2).
Protocol for Privacy-Respecting Distributed Event Detection.
Stefano Bennati, Catholijn Jonker. 2016
Presented at the 2016 Conference on Complex Systems (CCS).
The Role of Information in Group Formation.
Stefano Bennati, Leonard Wossnig, Johannes Thiele. 2016
Presented at the 2016 International Conference on Agents and Artificial Intelligence (ICAART).
On the Interaction between Learning and Evolution in Frequency-Dependent Foraging.
Stefano Bennati. 2016
Presented at the 2016 Computational Social Science Winter Symposium (CSSWS).
Self- vs. other-regarding behavior in zero-intelligence agents.
Stefano Bennati, Leonard Wossnig, Johannes Thiele, Dirk Helbing. 2015
Presented at the 2015 Computational Social Science Winter Symposium (CSSWS).
Human, model and machine: a complementary approach to big data.
Rob Thomson, Christian Lebiere, Stefano Bennati. 2014
Presented at the 2014 Workshop on Human Centered Big Data Research.
Extending the influence of contextual information in ACT-R using buffer decay.
Rob Thomson, Stefano Bennati, Christian Lebiere. 2014
Presented at the 2014 Annual meeting of the Cognitive Science Society (CogSci).
Functional cognitive models of malware identification.
C Lebiere, S Bennati, Rob Thomson, P Shakarian, E Nunes. 2015
Presented at the 2015 International Conference on Cognitive Modeling (ICCM).
Malware identification using cognitively-inspired inference.
R Thomson, C Lebiere, S Bennati, P Shakarian, E Nunes. 2015
Presented at the 2015 Conference on Behavior Representation in Modeling and Simulation (BRiMS).
Gestalt effects in planning: Rush-hour as an example.
S Bennati, S Brussow, M Ragni, L Konieczny. 2014
Presented at the 2014 Annual meeting of the Cognitive Science Society (CogSci).
Cognitive Robotics: Analysis of Preconditions and Implementation of a Cognitive Robotic System for Navigation Tasks.
Stefano Bennati, Marco Ragni. 2012
Presented at the 2012 International Conference on Cognitive Modeling (ICCM).

Working papers

PriMaL: A Privacy-Preserving Machine Learning Method for Event Detection in Distributed Sensor Networks.
Stefano Bennati, Catholijn Jonker. 2017

Published in: IN PRESS


Keywords: privacy, machine learning, distributed sensor networks, event detection


This paper introduces PriMaL, a general PRIvacy-preserving MAchine-Learning
method for reducing the privacy cost of information transmitted through a network. Distributed sensor networks are often used for automated classification and
detection of abnormal events in high-stakes situations, e.g. fire in buildings, earthquakes, or crowd disasters. Such networks might transmit privacy-sensitive information, e.g. GPS location of smartphones, which might be disclosed if the network
is compromised. Privacy concerns might slow down the adoption of the technology, in particular in the scenario of social sensing where participation is voluntary,
thus solutions are needed which improve privacy without compromising on the event
detection accuracy.
PriMaL is implemented as a machine-learning layer that works on top of an
existing event detection algorithm. Experiments are run in a general simulation
framework, for several network topologies and parameter values. The privacy footprint of state-of-the-art event detection algorithms is compared within the proposed
framework. Results show that PriMaL is able to reduce the privacy cost of a distributed event detection algorithm below that of the corresponding centralized algorithm, within the bounds of some assumptions about the protocol. Moreover the
performance of the distributed algorithm is not statistically worse than that of the
centralized algorithm.