Pattern Recognition Systems under Attack: Design Issues and Research Challenges

By Battista Biggio, Giorgio Fumera, Fabio Roli. 2014

92 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Computer Vision and Pattern Recognition, Safety, Robustness & Interpretability, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

In Computer Vision and Pattern Recognition

In Safety, Robustness & Interpretability

In Signal Processing

Citations in the corpus

Distillation as a Defense to Adversarial Perturbations Against Deep Neural Networks (2015, background): “Such attacks commonly occur today in non-DL classification systems [12], [13], [14], [15], [16].”
Wild Patterns: Ten Years After the Rise of Adversarial Machine Learning (2017, methodology): “Re-active approaches include: ( i ) timely detection of novel attacks, ( ii ) frequent classiﬁer retraining, and ( iii ) veriﬁcation of consistency of classiﬁer decisions against training data and ground-truth labels [40, 49].”
The Limitations of Deep Learning in Adversarial Settings (2015, background): “today on non-DL classification systems [5], [6], [21].”
Adversarial Detection of Flash Malware: Limitations and Open Issues (2017, methodology): “In some applications, the attacker may also exploit feedback on the classiﬁer’s decisions to improve her knowledge of the system, and, more generally, her attack strategy [7], [10], [11], [19].”
Yes, Machine Learning Can Be More Secure! A Case Study on Android Malware Detection (2017, methodology): “advocated in [9], [10], does not necessarily require one to trade classification accuracy in the absence of carefullycrafted attacks for improving security.”
Fast Feature Fool: A data independent approach to universal adversarial perturbations (2017, background): “Machine learning systems are vulnerable [2, 3, 9] to adversarial samples - malicious input with structured perturbations that can fool the systems to infer wrong predictions.”
DeepFense: Online Accelerated Defense Against Adversarial Deep Learning (2017, background): “at are reliable and trustworthy [1]. The existence of adversarial samples and their severe impact on the integrity of autonomous systems have been shown in the literature for both shallow [14], [40], [41], [42], [43] and deep [11], [12], [13], [44] learning models. Depending on the adversary goals, ”
Secure Kernel Machines against Evasion Attacks (2016, methodology): “To analyze possible attacks against machine learning and devise principled countermeasures in a more systematic manner, a formal model of the attacker has been proposed in [1, 4, 6, 7, 15].”
Ask, Acquire, and Attack: Data-free UAP Generation using Class Impressions (2018, background): “Machine learning models are pregnable (e.g. [4,3,9]) at test time to specially learned, mild noise in the input space, commonly known as adversarial perturbations.”
On the effectiveness of system API-related information for Android ransomware detection (2018, background): “odi- cations to the features of the analyzed test samples. However, this strategy may be challenging to be performed in practice. The problem, also known in the literature as Inverse Feature Mapping [31, 32, 33], is constructing the real sample that implements the modications made to the feature vec”
Towards Adversarial Malware Detection (2018, background): “) PDF malware samples. This issue has been widely known in the literature of adversarial machine learning, especially for optimization-based attacks, under the name of inverse feature-mapping problem [8, 12, 13, 44]. This problem arises from the fact that most of the optimization-based attacks propo”