Guenole C. M. Silvestre

Collaborative recommendation: A robustness analysis

Collaborative recommendation has emerged as an effective technique for personalized information access. However, there has been relatively little theoretical analysis of the conditions under which the technique is effective. To explore this issue, we analyse the <i>robustness</i> of collaborative recommendation: the ability to make recommendations despite (possibly intentional) noisy product ratings. There are two aspects to robustness: recommendation <i>accuracy</i> and <i>stability</i>. We formalize recommendation accuracy in machine learning terms and develop theoretically justified models of accuracy. In addition, we present a framework to examine recommendation stability in the context of a widely-used collaborative filtering algorithm. For each case, we evaluate our analysis using several real-world data-sets. Our investigation is both practically relevant for enterprises wondering whether collaborative recommendation leaves their marketing operations open to attack, and theoretically interesting for the light it sheds on a comprehensive theory of collaborative recommendation.

Detecting noise in recommender system databases

In this paper, we propose a framework that enables the detection of noise in recommender system databases. We consider two classes of noise: natural and malicious noise. The issue of natural noise arises from imperfect user behaviour (e.g. erroneous/careless preference selection) and the various rating collection processes that are employed. Malicious noise concerns the deliberate attempt to bias system output in some particular manner. We argue that both classes of noise are important and can adversely effect recommendation performance. Our objective is to devise techniques that enable system administrators to identify and remove from the recommendation process any such noise that is present in the data. We provide an empirical evaluation of our approach and demonstrate that it is successful with respect to key performance indicators.

Light degradation and voltage drift in polymer light-emitting diodes

It is shown that the voltage drift and light degradation in polymer light-emitting diodes are related and can be explained by the formation of traps and the modification of the space charge in the bulk of the polymer. The energy released by nonradiative carrier recombination is believed to be the driving force for the generation of traps in poly(p-phenylene vinylene) conjugated polymers. A first-approximation model is derived for the voltage drift and the light decrease during operation, which is in good agreement with experimental observations for time and current density dependencies.

Performance Analysis of Robust Audio Hashing

We present a novel theoretical analysis of the Philips audio fingerprinting method proposed by Haitsma, Kalker, and Oostveen (2001). Although this robust hashing algorithm exhibits very good performance, the method has only been partially analyzed in the literature. Hence, there is a clear need for a more complete analysis which allows both performance prediction and systematic optimization. We examine here the theoretical performance of the method for Gaussian inputs by means of a statistical model. Our analysis relies on formulating the unquantized fingerprint as a quadratic form, which affords a systematic way to compute the model parameters. We provide closed-form analytical upperbounds for the probability of bit error of the hash for two relevant scenarios: noise addition and desynchronization. We show that these results are useful when applied to real audio signals

JANIS: Just Another n-order Side-Informed Watermarking Scheme

This paper deals with some detection issues of watermark signals. We propose an easy way to implement an informed watermarking embedder whatever the detection function. This method shows that a linear detection function is not suitable for side information. This is the reason why we build a family of nonlinear functions named JANIS. Used with a side-informed embedder, its performance is much better than the classical spread spectrum method.

Utility-based neighbourhood formation for efficient and robust collaborative filtering

In this paper we propose novel neighbourhood formation and similarity weight transformation schemes for automated collaborative filtering systems. We demonstrate the benefits of our schemes from the point-of-view of the efficiency and robustness provided, while achieving the accuracy and coverage of a benchmark k-Nearest Neighbour (k-NN) model.

Joint iterative decoding and estimation for side-informed data hiding

We present a previously unavailable study on a general procedure for joint iterative decoding and estimation of attack parameters in side-informed data hiding. This type of approach, which exploits iteratively decodable codes for channel identification purposes, has recently become a relevant research trend in many digital communications problems. An advantage is that estimation pilots are not strictly required, thus affording in principle the implementation of blind methods that are able to work close to the theoretically maximum achievable rate. Such a target naturally requires the use of both near-optimum side-informed data hiding methods (e.g., DC-DM) and near-optimum iteratively decodable channel codes (e.g., turbo codes). The attack channels considered in this study are additive independent random noise, amplitude scaling, and a particular case of fine desynchronization of the sampling grid, whose parameters are estimated by maximum likelihood at the decoder. The complexity of this task is tackled by means of the Expectation-Maximization (EM) algorithm, relying on the use of a priori probabilities produced by the iterative decoding process.

Attacking Recommender Systems: A Cost-Benefit Analysis

A work highlights the lack of robustness collaborative recommender systems exhibit against attack. This vulnerability can lead to significantly biased recommendations for target items. Here, we examine such attacks from a cost perspective, focusing on how attack size - that is, the number of ratings inserted - affects attack success. We introduce a framework for quantifying the gains attackers realize, taking into account the financial cost of mounting the attack. A cost-benefit analysis of third-party attacks on recommender systems shows that attackers realize profits even when incurring costs associated with rating insertions.

An Evaluation of Neighbourhood Formation on the Performance of Collaborative Filtering

Personalisation features are key to the success of many web applications and collaborative recommender systems have been widely implemented. These systems assist users in finding relevant information or products from the vast quantities that are frequently available. In previous work, we have demonstrated that such systems are vulnerable to attack and that recommendations can be manipulated. We introduced the concept of robustness as a performance measure, which is defined as the ability of a system to provide consistent predictions in the presence of noise in the data. In this paper, we expand on our previous work by examining the effects of several neighbourhood formation schemes and similarity measures on system performance. We propose a neighbourhood filtering mechanism for filtering false profiles from the neighbourhood in order to improve the robustness of the system.

Embedding data in digital images using CDMA techniques

A watermarking technique is proposed for embedding a significant amount of data in digital still images while retaining a high perceptual quality. The scheme employs digital communication techniques to achieve high robustness to standard image processing operations. Information is embedded in the frequency domain by modulating selected DFT values of the image and using sets of orthogonal codes in a fashion similar to CDMA. DFT values are grouped into different bands defining independent channels for carrying data. Blind recovery of the embedded data is achieved by analyzing the DFT values of the watermarked image. The performance of the proposed adaptive scheme is evaluated by computer simulations.