ANTIDOTE
Published 4 November 2009
Benjamin I. P. Rubinstein, Blaine Nelson, Ling Huang, Anthony D. Joseph, Shing-hon Lau, Satish Rao
Citations344
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
This work proposes an antidote based on techniques from robust statistics and presents a new robust PCA-based detector that substantially reduces the effectiveness of poisoning for a variety of scenarios and indeed maintains a significantly better balance between false positives and false negatives than the original method when under attack.
Abstract
Statistical machine learning techniques have recently garnered increased popularity as a means to improve network design and security. For intrusion detection, such methods build a model for normal behavior from training data and detect attacks as deviations from that model. This process invites adversaries to manipulate the training data so that the learned model fails to detect subsequent attacks.
Keywords
Computer Science
Cambridge University Press eBooksPrediction, Learning, and Games
3,258 Citations2006Nicolò Cesa‐Bianchi, Gábor Lugosi
This chapter discusses prediction with expert advice, efficient forecasters for large classes of experts, and randomized prediction for specific losses.
TechnometricsControl Procedures for Residuals Associated With Principal Component Analysis
1,151 Citations1979J. Edward Jackson, Govind S. Mudholkar
Mining anomalies using traffic feature distributions
1,058 Citations2005Anukool Lakhina, Mark Crovella +1 more
It is argued that the distributions of packet features observed in flow traces reveals both the presence and the structure of a wide range of anomalies, and that using feature distributions, anomalies naturally fall into distinct and meaningful clusters that can be used to automatically classify anomalies and to uncover new anomaly types.
Diagnosing network-wide traffic anomalies
1,038 Citations2004Anukool Lakhina, Mark Crovella +1 more
A general method based on a separation of the high-dimensional space occupied by a set of network traffic measurements into disjoint subspaces corresponding to normal and anomalous network conditions to diagnose anomalies is proposed.
Adversarial classification
893 Citations2004Nilesh Dalvi, Pedro Domingos +3 more
This paper views classification as a game between the classifier and the adversary, and produces a classifier that is optimal given the adversary's optimal strategy, and experiments show that this approach can greatly outperform a classifiers learned in the standard way.
Can machine learning be secure?
863 Citations2006Marco Barreno, Blaine Nelson +3 more
A taxonomy of different types of attacks on machine learning techniques and systems, a variety of defenses against those attacks, and an analytical model giving a lower bound on attacker's work function are provided.
Adversarial learning
731 Citations2005Daniel Lowd, Christopher Meek
This paper introduces the adversarial classifier reverse engineering (ACRE) learning problem, the task of learning sufficient information about a classifier to construct adversarial attacks, and presents efficient algorithms for reverse engineering linear classifiers with either continuous or Boolean features.
Polygraph: Automatically Generating Signatures for Polymorphic Worms
675 Citations2005J.P. Newsome, Brad Karp +1 more
This paper presents Polygraph, a signature generation system that successfully produces signatures that match polymorphic worms, and proposes classes of signature suited for matching polymorphic worm payloads and presents algorithms for automatic generation of signatures in these classes.
Characterization of network-wide anomalies in traffic flows
412 Citations2004Anukool Lakhina, Mark Crovella +1 more
This paper presents the first large-scale exploration of the power of the subspace method when applied to flow traffic, and finds that almost all of the anomalies detected represent events of interest to network operators.
Towards highly reliable enterprise network services via inference of multi-level dependencies
373 Citations2007Paramvir Bahl, Ranveer Chandra +4 more
An Inference Graph model is introduced, which is well-adapted to user-perceptible problems rooted in conditions giving rise to both partial service degradation and hard faults, and takes into account multi-level structure, which leads to a 30% improvement in fault localization, as compared to two-level approaches.
Journal of the American Statistical AssociationRobust Estimation of Dispersion Matrices and Principal Components
373 Citations1981Susan J. Devlin, R. Gnanadesikan +1 more
Sensitivity of PCA for traffic anomaly detection
351 Citations2007Haakon Ringberg, Augustin Soule +2 more
This study identifies and evaluates four main challenges of using PCA to detect traffic anomalies: the false positive rate is very sensitive to small differences in the number of principal components in the normal subspace, the effectiveness of PCA is sensitive to the level of aggregation of the traffic measurements, a large anomaly may in advertently pollute the normalSubspace.
Journal of Multivariate AnalysisHigh breakdown estimators for principal components: the projection-pursuit approach revisited
333 Citations2004Christophe Croux, Anne Ruiz‐Gazen
Edinburgh Research Explorer (University of Edinburgh)Exploiting machine learning to subvert your spam filter
325 Citations2008Blaine Nelson, Marco Barreno +7 more
This paper shows how an adversary can exploit statistical machine learning, as used in the SpamBayes spam filter, to render it useless--even if the adversary's access is limited to only 1% of the training messages.
Journal of the American Statistical AssociationProjection-Pursuit Approach to Robust Dispersion Matrices and Principal Components: Primary Theory and Monte Carlo
290 Citations1985Guoying Li, Zhonglian Chen
Chemometrics and Intelligent Laboratory SystemsAlgorithms for Projection–Pursuit robust principal component analysis
270 Citations2007Christophe Croux, Peter Filzmoser +1 more
Filtering spam with behavioral blacklisting
267 Citations2007Anirudh Ramachandran, Nick Feamster +1 more
SpamTracker is a spam filtering system that uses a new technique called behavioral blacklisting to classify email senders based on their sending behavior rather than their identity, and it is found that SpamTracker can correctly classify many spammers missed by current filtering techniques.
Detection and identification of network anomalies using sketch subspaces
228 Citations2006Xin Li, Fang Bian +5 more
This paper shows how to combine traffic sketches with a subspace method to detect anomalies with high accuracy and identify the IP flows(s) that are responsible for the anomaly.
Lecture notes in computer scienceParagraph: Thwarting Signature Learning by Training Maliciously
199 Citations2006James Newsome, Brad Karp +1 more
It is shown that even a delusive adversary, whose samples are all correctly labeled, can obstruct learning, and practical attacks against learning are described, in which an adversary constructs labeled samples that prevent or severely delay generation of an accurate classifier.
The MIT Press eBooksIn-Network PCA and Anomaly Detection
195 Citations2007Ling Huang, XuanLong Nguyen +4 more
SMURF is proposed, the first declarative, adaptive smoothing filter for RFID data cleaning, which models the unreliability of RFID readings by viewing RFID streams as a statistical sample of tags in the physical world, and exploits techniques grounded in sampling theory to drive its cleaning processes.
TechnometricsPrincipal Components and Orthogonal Regression Based on Robust Scales
193 Citations2005Ricardo A. Maronna
Comparisons based on both simulated and real data show that the proposed procedure is more robust than its competitors, especially for large m.
On Attacking Statistical Spam Filters.
173 Citations2004Gregory L. Wittel, Shyhtsun Felix Wu
The general attack methods spammers use are examined, along with challenges faced by developers and spammers, and an attack is demonstrated that attempts to more strongly work against the statistical nature behind spam systems.
Applying PCA for Traffic Anomaly Detection: Problems and Solutions
172 Citations2009Daniela Brauckhoff, Kavé Salamatian +1 more
A slightly modified version of PCA is developed that uses only data from a single router and proposes a solution to deal with the main problem, that PCA fails to capture temporal correlation, and is replaced with the Karhunen-Loeve transform.
Evading network anomaly detection systems
158 Citations2006Prahlad Fogla, Wenke Lee
This paper presents a formal framework for the open problem: given an anomaly detection system and an attack, can one automatically generate its PBA instances and suggests how the IDS can be improved to prevent the PBA.
Automating cross-layer diagnosis of enterprise wireless networks
117 Citations2007Yuchung Cheng, Mikhail Afanasyev +6 more
This paper focuses on data transfer delays unique to 802.11 networks - media access dynamics and mobility management latency and reconstructs sources of delay - from the physical layer to the transport - layer as well as the interactions among them.
Journal of nonparametric statisticsGeneralizing univariate signed rank statistics for testing and estimating a multivariate location parameter
101 Citations1995Ola Hössjer, Christophe Croux
ACM SIGCOMM Computer Communication ReviewWhat's going on?
84 Citations2008Srikanth Kandula, Ranveer Chandra +1 more
What's going on?
66 Citations2008Srikanth Kandula, Ranveer Chandra +1 more
From packet timing information, eXpose learns rules for network communication that may be spread across multiple hosts, protocols or applications that help with network monitoring, diagnosis, and intrusion detection with few false positives.
ACM SIGMETRICS Performance Evaluation ReviewStealthy poisoning attacks on PCA-based anomaly detectors
62 Citations2009Benjamin I. P. Rubinstein, Blaine Nelson +6 more
ROC curves provide a comprehensive analysis of PCA-based detection on contaminated data, and show that even small attacks can undermine this otherwise successful anomaly detector.
SSRN Electronic JournalAlgorithms for Projection-Pursuit Robust Principal Component Analysis
55 Citations2006Christophe Croux, Peter Filzmoser +1 more
It is shown that the currently available algorithm for robust Projection-Pursuit PCA performs poor in presence of many variables and a new algorithm is proposed that is more suitable for the analysis of chemical data.
FigshareLimits of Learning-Based Signature Generation with Adversaries
52 Citations2018Shobha Venkataraman, Avrim Blum +1 more
This paper forms a framework that allows a unified analysis of pattern-extraction algorithms for signaturegeneration in an adversarial setting, and proves lower bounds on the number of mistakes any patternextraction learning algorithmmust make under common assumptions, by showing how to adapt results from learning theory.
Detecting Distributed Attacks using Network-Wide Flow Traffic
13 Citations2005Anukool Lakhina, Mark Crovella +1 more
The more distributed an attack is, the better the methods are at detecting it, and it is demonstrated that attacks rates of less than 1% of the underlying traffic can be detected successfully by the methods.