Publications‎ > ‎

International Conference Publications

Evolutionary Synthesis of Sensing Controllers for Voxel-based Soft Robots

posted May 2, 2019, 5:00 AM by Eric Medvet   [ updated May 2, 2019, 5:00 AM ]

  • Annual Conference on Artificial Life (ALIFE), 2019, Newcastle upon Tyne (United Kingdom), to appear
  • Jacopo Talamini, Eric Medvet, Alberto Bartoli, Andrea De Lorenzo
Soft robots allow for interesting morphological and behavioral designs because they exhibit more degrees of freedom than robots composed of rigid parts. In particular, voxel-based soft robots (VSRs)—aggregations of elastic cubic building blocks—have attracted the interest of Robotics and Artificial Life researchers. VSRs can be controlled by changing the volume of individual blocks: simple, yet effective controllers that do not exploit the feedback of the environment, have been automatically designed by means of Evolutionary Algorithms (EAs).
In this work we explore the possibility of evolving sensing controllers in the form of artificial neural networks: we hence allow the robot to sense the environment in which it moves. Although the search space for a sensing controller is larger than its non-sensing counterpart, we show that effective sensing controllers can be evolved which realize interesting locomotion behaviors. We also experimentally investigate the impact of the VSR morphology on the effectiveness of the search and verify that the sensing controllers are indeed able to exploit their sensing ability for better solving the locomotion task.

An Analysis of Dimensionality Reduction Techniques for Visualizing Evolution

posted Apr 18, 2019, 6:35 AM by Eric Medvet   [ updated Apr 23, 2019, 3:31 AM ]

  • 10th annual workshop on Visualisation in Genetic and Evolutionary Computation (VizGEC), 2019, Prague (Czech Republic), to appear
  • Andrea De Lorenzo, Eric Medvet, Tea Tušar, Alberto Bartoli
We consider the problem of visualizing the population dynamics along an evolutionary run using a dimensionality reduction technique for mapping individuals from the original search space to a 2-D space. We quantitatively assess four of these techniques in terms of their ability to preserve useful information about (a) population movements and (b) exploration-exploitation trade-off. We propose two compact visualizations aimed at highlighting these two aspects of population dynamics and evaluate them qualitatively. The results are very promising as the proposed framework is indeed able to represent crucial properties of population dynamics in a way that is both highly informative and simple to understand.

Design of Powered Floor Systems for Mobile Robots with Differential Evolution

posted Jan 7, 2019, 7:16 AM by Eric Medvet   [ updated Apr 15, 2019, 1:10 AM ]

Mobile robots depend on power for performing their task. Powered floor systems, i.e., surfaces with conductive strips alternatively connected to the two poles of a power source, are a practical and effective way for supplying power to robots without interruptions, by means of sliding contacts. Deciding where to place the sliding contacts so as to guarantee that a robot is actually powered irrespective of its position and orientation is a difficult task. We here propose a solution based on Differential Evolution: we formally define problem-specific constraints and objectives and we use them for driving the evolutionary search. We validate experimentally our proposed solution by applying it to three real robots and by studying the impact of the main problem parameters on the effectiveness of the evolved designs for the sliding contacts. The experimental results suggest that our solution may be useful in practice for assisting the design of powered floor systems.

Communication-based Cooperative Tasks: how the Language Expressiveness affects Reinforcement Learning

posted Nov 26, 2018, 6:29 AM by Eric Medvet   [ updated May 13, 2019, 6:47 AM ]

We consider a cooperative multi-agent system in which cooperation may be enforced by communication between agents but in which agents must learn to communicate. The system consists of a game in which agents may move in a 2D world and are given the task of reaching specified targets. Each agent knows the target of another agent but not its own, thus the only way to solve the task is for the agents to guide one another using communication and, in particular, by learning how to communicate. We cast this game in terms of a partially observed Markov game and show that agents may learn policies for moving and communicating in the form of a neural network by means of reinforcement learning. We investigate in depth the impact on the learning quality of the expressiveness of the language, which is a function of vocabulary size, number of agents and number of targets.

Back To The Basics: Security of Software Downloads for Smart Objects

posted Oct 8, 2018, 3:03 AM by Eric Medvet   [ updated Jan 11, 2019, 5:53 AM ]

Smart objects will soon pervade our homes, cities, factories, plants, and hospitals and this fact will introduce widespread important risks for the society as a whole, due to unavoidable security vulnerabilities of those objects. The problem of updating the software of smart objects in order to fix vulnerabilities will thus become of crucial importance. In this work we investigate the security of current software download environments for smart objects. This investigation allows gaining important insights into the security awareness of organizations that distribute software across the web and, more broadly, on their readiness to take control of our everyday life.

Personalized, Browser-based Visual Phishing Detection Based on Deep Learning

posted Sep 5, 2018, 3:48 AM by Eric Medvet   [ updated Jan 27, 2019, 11:44 PM ]

Phishing defense mechanisms that are close to browsers and that do not rely on any forms of website reputation may be a powerful tool for combating phishing campaigns that are increasingly more targeted and last for increasingly shorter life spans. Browser-based phishing detectors that are specialized for a user-selected set of targeted web sites and that are based only on the overall visual appearance of a target could be a very effective tool in this respect. Approaches of this kind have not been very successful for several reasons, including the difficulty of coping with the large set of genuine pages encountered in normal browser usage without flooding the user with false positives. In this work we intend to investigate whether then the power of modern deep learning methodologies for image classification may enable solutions that are more practical and effective. Our experimental assessment of a convolutional neural network resulted in very high classification accuracy for targeted sets of 15 websites (the largest size that we analyzed) even when immersed in a set of login pages taken from 100 websites.

Observing the Population Dynamics in GE by means of the Intrinsic Dimension

posted Jul 4, 2018, 9:15 AM by Eric Medvet   [ updated Dec 7, 2018, 12:55 AM ]

  • Evolutionary Machine Learning workshop at International Conference on Parallel Problem Solving from Nature (EML@PPSN), 2018, Coimbra (Portugal)
  • Eric Medvet, Alberto Bartoli, Alessio Ansuini, Fabiano Tarlao
  • Google Scholar, arXiv
We explore the use of Intrinsic Dimension (ID) for gaining insights in how populations evolve in Evolutionary Algorithms. ID measures the minimum number of dimensions needed to accurately describe a dataset and its estimators are being used more and more in Machine Learning to cope with large datasets. We postulate that ID can provide information about population which is complimentary w.r.t. what (a simple measure of) diversity tells. We experimented with the application of ID to populations evolved with a recent variant of Grammatical Evolution. The preliminary results suggest that diversity and ID constitute two different points of view on the population dynamics.

Detection of Obfuscation Techniques in Android Applications

posted Jun 11, 2018, 1:17 AM by Eric Medvet   [ updated Sep 11, 2018, 12:11 AM ]

Current signature detection mechanisms can be easily evaded by malware writers by applying obfuscation techniques. Employing morphing code techniques, attackers are able to generate several variants of one malicious sample, making the corresponding signature obsolete. Considering that the signature definition is a laborious process manually performed by security analysts, in this paper we propose a method, exploiting static analysis and Machine Learning classification algorithms, to identify whether a mobile application is modified by means of one or more morphing techniques. We perform experiments on a real-world dataset of Android applications (morphed and original), obtaining encouraging results in the obfuscation technique(s) identification.

(In)Secure Configuration Practices of WPA2 Enterprise Supplicants

posted Jun 11, 2018, 1:12 AM by Eric Medvet   [ updated Sep 11, 2018, 12:11 AM ]

WPA2 Enterprise is a fundamental technology for secure communication in enterprise wireless networks. A key requirement of this technology is that WiFi-enabled devices (i.e., supplicants) be correctly configured before connecting to the enterprise wireless network. Supplicants that are not configured correctly may fall prey of attacks aimed at stealing the network credentials very easily. Such credentials have an enormous value because they usually unlock access to all enterprise services.
In this work we investigate whether users and technicians are aware of these important and widespread risks. We conducted two extensive analyses: a survey among approximately 1000 users about how they configured their WiFi devices for enterprise network access; and, a review of approximately 310 network configuration guides made available by enterprise network administrators. The results provide strong indications that the key requirement of WPA2 Enterprise is violated systematically and thus can no longer be considered realistic.

GOMGE: Gene-pool Optimal Mixing on Grammatical Evolution

posted May 15, 2018, 5:41 AM by Eric Medvet   [ updated Sep 10, 2018, 7:31 AM ]

Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA) is a recent Evolutionary Algorithm (EA) in which the interactions among parts of the solution (i.e., the linkage) are learned and exploited in a novel variation operator. We present GOMGE, the extension of GOMEA to Grammatical Evolution (GE), a popular EA based on an indirect representation which may be applied to any problem whose solutions can be described using a context-free grammar (CFG). GE is a general approach that does not require the user to tune the internals of the EA to fit the problem at hand: there is hence the opportunity for benefiting from the potential of GOMEA to automatically learn and exploit the linkage. We apply the proposed approach to three variants of GE differing in the representation (original GE, SGE, and WHGE) and incorporate in GOMGE two specific improvements aimed at coping with the high degeneracy of those representations. We experimentally assess GOMGE and show that, when coupled with WHGE and SGE, it is clearly beneficial to both effectiveness and efficiency, whereas it delivers mixed results with the original GE.

1-10 of 73