A5065359946- Biometric Identification and Security
- Adversarial Robustness in Machine Learning
- Advanced Malware Detection Techniques
- Face and Expression Recognition
- Anomaly Detection Techniques and Applications
- Network Security and Intrusion Detection
- Neural Networks and Applications
- Face recognition and analysis
- User Authentication and Security Systems
- Forensic Fingerprint Detection Methods
- Remote-Sensing Image Classification
- Spam and Phishing Detection
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Digital Media Forensic Detection
- Machine Learning and Data Classification
- Evolutionary Algorithms and Applications
- Fuzzy Logic and Control Systems
- Forensic and Genetic Research
- Image Retrieval and Classification Techniques
- Human Pose and Action Recognition
- Internet Traffic Analysis and Secure E-voting
- Spectroscopy and Chemometric Analyses
- Advanced Steganography and Watermarking Techniques
- Text and Document Classification Technologies
University of Cagliari
2015-2025
University of Genoa
1996-2025
Northwestern Polytechnical University
2020-2023
Institute of Electrical and Electronics Engineers
2014-2022
École Polytechnique Fédérale de Lausanne
2022
Ca' Foscari University of Venice
2022
Regional Municipality of Niagara
2022
IEEE Computer Society
2022
University of Namur
2022
Institut national de recherche en informatique et en automatique
2022
A number of online services nowadays rely upon machine learning to extract valuable information from data collected in the wild. This exposes algorithms threat poisoning, i.e., a coordinate attack which fraction training is controlled by attacker and manipulated subvert process. To date, these attacks have been devised only against limited class binary algorithms, due inherent complexity gradient-based procedure used optimize poisoning points (a.k.a. adversarial examples). In this work, we...
Pattern classification systems are commonly used in adversarial applications, like biometric authentication, network intrusion detection, and spam filtering, which data can be purposely manipulated by humans to undermine their operation. As this scenario is not taken into account classical design methods, pattern may exhibit vulnerabilities, whose exploitation severely affect performance, consequently limit practical utility. Extending theory methods settings thus a novel very relevant...
Machine learning has already been exploited as a useful tool for detecting malicious executable files. Data retrieved from malware samples, such header fields, instruction sequences, or even raw bytes, is leveraged to learn models that discriminate between benign and software. However, it also shown machine deep neural networks can be fooled by evasion attacks (also known adversarial examples), i.e., small changes the input data cause misclassification at test time. In this work, we...
Learning in adversarial settings is becoming an important task for application domains where attackers may inject malicious data into the training set to subvert normal operation of data-driven technologies. Feature selection has been widely used machine learning security applications improve generalization and computational efficiency, although it not clear whether its use be beneficial or even counterproductive when are poisoned by intelligent attackers. In this work, we shed light on...
To cope with the increasing variability and sophistication of modern attacks, machine learning has been widely adopted as a statistically-sound tool for malware detection. However, its security against well-crafted attacks not only recently questioned, but it shown that exhibits inherent vulnerabilities can be exploited to evade detection at test time. In other words, itself weakest link in system. this paper, we rely upon previously-proposed attack framework categorize potential scenarios...
Pattern recognition and machine learning techniques have been increasingly adopted in adversarial settings such as spam, intrusion, malware detection, although their security against well-crafted attacks that aim to evade detection by manipulating data at test time has not yet thoroughly assessed. While previous work mainly focused on devising adversary-aware classification algorithms counter evasion attempts, only few authors considered the impact of using reduced feature sets classifier...
Undersampling is a widely adopted method to deal with imbalance pattern classification problems. Current methods mainly depend on either random resampling the majority class or at decision boundary. Random-based undersampling fails take into consideration informative samples in data while boundary sensitive overlapping. Both techniques ignore distribution information of training dataset. In this paper, we propose diversified sensitivity-based method. Samples are clustered capture and enhance...
Windows malware detectors based on machine learning are vulnerable to adversarial examples, even if the attacker is only given black-box query access model. The main drawback of these attacks that: (i) they query-inefficient, as rely iteratively applying random transformations input malware; and (ii) may also require executing in a sandbox at each iteration optimization process, ensure that its intrusive functionality preserved. In this paper, we overcome issues by presenting novel family...
In this paper, a theoretical and experimental analysis of linear combiners for multiple classifier systems is presented. Although are the most frequently used combining rules, many important issues related to their operation pattern classification tasks lack basis. After critical review framework developed in works by Tumer Ghosh on which our based, we focus simplest widely implementation combiners, consists assigning nonnegative weight each individual classifier. Moreover, consider ideal...
The problem of extending the Jeffreys-Matusita distance to multiclass cases for feature-selection purposes is addressed and a solution equivalent Bhattacharyya bound presented. This extension compared with widely used weighted average both by examining respective formulae experimenting on an optical remote-sensing data set.
A spoof or fake is a counterfeit biometric that used in an attempt to circumvent sensor Liveness detection distinguishes between live and traits. based on the principle additional information can be garnered above beyond data procured by standard verification system, this verify if measure authentic. The Fingerprint Detection Competition (LivDet) goal compare both software-based (Part 1) hardware-based 2) fingerprint liveness methodologies open all academic industrial institutions....
"Liveness detection", a technique used to determine the vitality of submitted biometric, has been implemented in fingerprint scanners recent years. The goal for LivDet 2011 competition is compare software-based liveness detection methodologies (Part 1), as well systems which incorporate capabilities 2), using standardized testing protocol and large quantities spoof live images. This was open all academic industrial institutions have solution either or system-based problem. Five submissions...