Publications

Refine Results

(Filters Applied) Clear All

Deep Neural Network Approaches to Speaker and Language Recognition(323.6 KB)

Date:
October 1, 2015
Published in:
IEEE Signal Processing Letters, vol. 22, no. 10
Type:
Journal Article

Summary

The impressive gains in performance obtained using deep neural networks (DNNs) for automatic speech recognition (ASR) have motivated the application of DNNs to other speech technologies such as speaker recognition (SR) and language recognition (LR). In this work we present the application of single DNN for both SR and LR using the 2013 Domain Adaptation Challenge speaker recognition (DAC13)and the NIST 2011 language recognition evaluation (LRE11) benchmarks.

The AFRL-MITLL WMT15 System: There's More Than One Way To Decode It!(318.6 KB)

Date:
September 17, 2015
Published in:
Proceedings of the 10th Workshop on Machine Translation (WMT '15)
Type:
Conference Paper

Summary

This paper describes the AFRL-MITLL statistical MT systems and the improvements that were developed during the WMT15 evaluation campaign. As part of these efforts we experimented with a number of extensions to the standard phrase-based model that improve performance on the Russian to English translation task creating three submission systems with different decoding strategies.

Characterizing Phishing Threats with Natural Language Processing(404.72 KB)

Author:
Date:
September 15, 2015
Published in:
Proceedings of the 2015 IEEE Conference on Communications and Network Security (CNS) , Florence, Italy
Type:
Conference Paper

Summary

Spear phishing is a widespread concern in the modern network security landscape, but there are few metrics that measure the extent to which reconnaissance is performed on phishing targets. Spear phishing emails closely match the expectations of the recipient, based on details of their experiences and interests, making them a popular propagation vector for harmful malware. In this work we use Natural Language Processing techniques to investigate a specific real-world phishing campaign and quantify attributes that indicate a targeted spear phishing attack.

A Unified Deep Neural Network for Speaker and Language Recognition(254.34 KB)

Date:
September 8, 2015
Published in:
Proceedings of Interspeech 2015, Dresden, Germany
Type:
Conference Paper

Summary

Significant performance gains have been reported separately for speaker recognition (SR) and language recognition (LR) tasks using either DNN posteriors of sub-phonetic units or DNN feature representations, but the two techniques have not been compared on the same SR or LR task or across SR and LR tasks using the same DNN. In this work we present the application of a single DNN for both tasks using the 2013 Domain Adaptation Challenge speaker recognition (DAC13) and the NIST 2011 language recognition evaluation (LRE11) benchmarks.

Unifying Leakage Classes: Simulatable Leakage and Pseudoentropy(324.96 KB)

Date:
September 4, 2015
Published in:
Information Theoretic Security, Lecture Notes in Computer Science, vol. 9063, pp. 69-86
Type:
Journal Article
Topic:

Summary

Leakage resilient cryptography designs systems to withstand partial adversary knowledge of secret state. Ideally, leakage-resilient systems withstand current and future attacks, restoring confidence in the security of implemented cryptographic systems. Understanding the relation between classes of leakage functions is an important aspect. In this work, we consider the memory leakage model, where the leakage class contains functions over the system’s entire secret state. Standard limitations include functions with bounded output length, functions that retain (pseudo) entropy in the secret, and functions that leave the secret computationally unpredictable.

A spectral framework for anomalous subgraph detection(2.65 MB)

Date:
August 15, 2015
Published in:
IEEE Transactions on Signal Processing, vol. 63, no. 16, pp. 4191–4206
Type:
Journal Article
Topic:

Summary

In numerous applications, the data of interest consist of entities and the relationships between them. A common problem is the detection of a subset of entities whose connectivity is anomalous with respect to the rest of the data. While the detection of such anomalous subgraphs has received a substantial amount of attention, no application-agnostic framework exists for analysis of signal detectability in graph-based data. In this paper, we describe a framework that enables such analysis using the principal eigenspace of a graph’s residuals matrix, commonly called the modularity matrix in community detection.

En Route Sector Capacity Model Final Report(1.98 MB)

Author:
Date:
August 15, 2015
Published in:
Project Report ATC-426, MIT Lincoln Laboratory
Type:
Project Report

Summary

Accurate predictions of en route sector capacity are vital when analyzing the benefits of proposed new air traffic management decision-support tools or new airspace designs. Controller workload is the main determinant of sector capacity. This report describes a new workload-based capacity model that improves upon the Federal Aviation Administration’s current Monitor Alert capacity model.

Iris Biometric Security Challenges and Possible Solutions: For your eyes only - Using the iris as a key(2.7 MB)

Date:
August 13, 2015
Published in:
Signal Processing Magazine, IEEE , vol. 32, no. 5, pp. 42-53
Type:
Journal Article
Topic:

Summary

In this article, we illustrate a metric that can be used to optimize biometrics for authentication. Using iris biometrics as an example, we explore possible directions for improving processing and representation according to this metric. Finally, we discuss why strong biometric authentication remains a challenging problem and propose some possible future directions for addressing these challenges.

Trustworthy whole-system provenance for the linux kernel(682.54 KB)

Date:
August 12, 2015
Published in:
24th USENIX Security Symposium (USENIX Security 15), Washington, D.C.
Type:
Conference Paper

Summary

A provenance-aware system automatically gathers and reports metadata that describes the history of each object being processed on the system. Provenance itself is a ripe attack vector, and its authenticity and integrity must be guaranteed before it can be put to use. We present Linux Provenance Modules (LPM), the first general framework for the development of provenance-aware systems.

Simulation based Evaluation of a Code Diversification Strategy(923.91 KB)

Date:
July 21, 2015
Published in:
Proceedings of the 5th International Conference on Simulation and Modeling Methodologies, Technologies and Applications, Colmar, Alsace, France
Type:
Conference Paper

Summary

Periodic randomization of a computer program’s binary code is an attractive technique for defending against several classes of advanced threats. In this paper we describe a model of attacker-defender interaction in which the defender employs such a technique against an attacker who is actively constructing an exploit using Return Oriented Programming (ROP).