Enabling Fast and Universal Audio Adversarial Attack Using Generative
Model
2
Citation
0
Reference
10
Related Paper
Citation Trend
Abstract:
Recently, the vulnerability of DNN-based audio systems to adversarial attacks has obtained the increasing attention. However, the existing audio adversarial attacks allow the adversary to possess the entire user's audio input as well as granting sufficient time budget to generate the adversarial perturbations. These idealized assumptions, however, makes the existing audio adversarial attacks mostly impossible to be launched in a timely fashion in practice (e.g., playing unnoticeable adversarial perturbations along with user's streaming input). To overcome these limitations, in this paper we propose fast audio adversarial perturbation generator (FAPG), which uses generative model to generate adversarial perturbations for the audio input in a single forward pass, thereby drastically improving the perturbation generation speed. Built on the top of FAPG, we further propose universal audio adversarial perturbation generator (UAPG), a scheme crafting universal adversarial perturbation that can be imposed on arbitrary benign audio input to cause misclassification. Extensive experiments show that our proposed FAPG can achieve up to 167X speedup over the state-of-the-art audio adversarial attack methods. Also our proposed UAPG can generate universal adversarial perturbation that achieves much better attack performance than the state-of-the-art solutions.Using recent improvements to Valiant’s algorithm for parsing contextfree languages, we present an implementation of a generator of parsers that works incrementally, that can be parallelized and generated from a grammar specification. Using a tree structure makes for both easy use of incrementality and parallelization. The resulting code is reasonably fast and handles correct input in a satisfactory way, and would be suitable for use in a text editor setting, where small changes are frequent but only should lead to minimal work.
Tree (set theory)
Code (set theory)
Cite
Citations (0)
Limiting
Cite
Citations (6)
Vetting
Adversarial machine learning
Sketch
Evasion (ethics)
Heuristics
Cite
Citations (1,717)
We have developed a new experimental method for interrogating statistical theories of music perception by implementing these theories as generative music algorithms. We call this method Generation in Context. This method differs from most experimental techniques in music perception in that it incorporates aesthetic judgments. Generation In Context is designed to measure percepts for which the musical context is suspected to play an important role. In particular the method is suitable for the study of perceptual parameters which are temporally dynamic. We outline a use of this approach to investigate David Temperley's (2007) probabilistic melody model, and provide some provisional insights as to what is revealed about the model. We suggest that Temperley's model could be improved by dynamically modulating the probability distributions according to the changing musical context.
Generative model
Cite
Citations (0)
Rapid advancement of our information society necessitates prompt update and expansion of the technical scope and focus of interest of our IEEE Signal Processing Society (SPS). Compared with just some years ago, the current focus of signal processing as an enabling technology has been significantly broadened. Now it encompasses theories, architectures, algorithms, implementations, and applications for the transformation of information contained in many different physical, symbolic, or abstract formats that we broadly designate as “signals.” Methodology wise, signal processing uses mathematical, statistical, computational, heuristic, and/ or linguistic representations, formalisms, and techniques for sensing, acquisition, extraction, representation, modeling, analysis, synthesis, compression, detection, recovery, decomposition, enhancement, rendering, display, learning, recognition, un derstanding, securing, authenticating, and communicating of information and signals. Such diverse “processing” tasks are accomplished by either digital or analog devices or algorithms, and in the form of either software, hardware, or firmware.
Firmware
Statistical signal processing
Cite
Citations (1)
A formal model of an in-circuit program generator based on a fact-algorithm-rule structure is described. The generalized structure of the model will allow modification of the system and its databases in directions developed by the user rather than the original designer. This approach will also provide a structure on which to base the combined in-circuit and functional test programs required in the near future.
Representation
Base (topology)
Digital pattern generator
Cite
Citations (0)
Presented here is a step by step analysis of how the basic Sparse Distributed Memory (SDM) model can be modified to enhance its generalization capabilities for classification tasks. Data is taken from speech generated by a single talker. Experiments are used to investigate the theory of associative memories and the question of generalization from specific instances.
Associative property
Cite
Citations (14)
Aggregating and sequencing of the content units is at the core of e-learning theories and standards. We discuss the aggregating/sequencing problems in the context of using generative learning objects (GLOs). Proposed by Boyle, Morales, Leeder in 2004, GLOs provide more capabilities, focus on quality issues, and introduce a solid basis for a marked improvement in productivity. We use meta-programming techniques to specify GLOs and then to automatically generate LO units on demand. Aggregating of the generated units to form a compound at a higher granularity level can be performed in various ways depending on the selected criteria or their trade-offs (e.g., complexity, granularity level, semantic density, time constraints, capabilities of modelling the learning process, etc.) that enable to evaluate units in advance. We describe aggregating as an internal sequencing of the content units derived from a GLO. Our contribution is a formal graph-based model to specify the problem when the variability of LO units is large. First we formulate the problem and consider properties of the proposed model; and then we analyze a case study, implementation capabilities, and evaluate the approach for e-learning.
Generative model
Learning object
Granularity
Cite
Citations (7)
Automatic speech understanding and automatic speech recognition extract different kinds of information from the input signal. The result of the former must be evaluated on the basis of the response of the system while the result of the latter is the word sequence which matches the input signal. In both cases search has to be performed based on scores of interpretation hypotheses.
A scoring method is presented based on stochastic context-free grammars. The method gives optimal upper-bounds for the computation of the best derivation trees of a sentence. This method allows language models to be built based on stochastic context-free grammars and their use with an admissible search algorithm that interprets a speech signal with left-to-right or middle-out strategies. Theoretical and computational aspects are discussed.
SIGNAL (programming language)
Sequence (biology)
Cite
Citations (7)
Computer programs for the game of Go play only at the level of an advanced beginning player. The standard approach to constructing a program based on brute force game-tree search does not work well because of the game tree size and, more significantly, the difficulty in constructing fast, accurate heuristic evaluation functions. In this paper, we consider the use of intent inference in a Go program. In particular, we discuss how models of an opponent’s long-term playing style and short-term intentions can direct the exploration of candidate moves and influence the evaluation of game positions. We propose a probabilistic approach to user modeling and intent inference, and we note key issues relevant to the implementation of an intent inference agent.
Game tree
Tree (set theory)
Backward chaining
Cite
Citations (0)