Authorship verification of e-mail and tweet messages applied for continuous authentication

Abstract

Authorship verification using stylometry consists of identifying a user based on his writing style. In this paper, authorship verification is applied for continuous authentication using unstructured online text-based entry. An online document is decomposed into consecutive blocks of short texts over which (continuous) authentication decisions happen, discriminating between legitimate and impostor behaviors. We investigate blocks of texts with 140, 280 and 500 characters. The feature set includes traditional features such as lexical, syntactic, application specific features, and new features extracted from n-gram analysis. Furthermore, the proposed approach includes a strategy to circumvent issues related to unbalanced dataset, and uses Information Gain and Mutual Information as a feature selection strategy and Support Vector Machine (SVM) for classification. Experimental evaluation of the proposed approach based on the Enron email and Twitter corpuses yields very promising results consisting of an Equal Error Rate (EER) varying from 9.98% to 21.45%, for different block sizes.

Authorship Verification: A Review of Recent Advances

“…The average performance of best systems in all PAN corpora (see Table 3) indicates that the error rate of state-of-the-art methods in authorship verification is around 20%. 7 Although this is too high in comparison to the most effective technologies used to provide forensic evidence (e.g., the error rate of DNA analysis is less than 1% [28]), it is comparable to other technologies that analyse noisy data, like latent fingerprint matching [12] or speaker identification [8]. The relatively higher AUROC scores indicate that the verification models are able to rank answers more effectively and they Fig.…”

“…When multiple labelled texts are available, they combine the answers to provide the final decision. Typical examples of this category are described by Seidman [55], Jankowska et al [22], Moreau et al [41], Brocardo et al [7], and Castro-Castro et al [9]. Another variation is to first concatenate all labelled texts and then split the resulting text into samples of equal size [5,17].…”

“…PAN evaluation campaigns provided benchmark corpora covering several natural languages and genres as well as an experimentation and evaluation framework to assess the performance of multiple verification methods. Since 2013, significant progress has been reported and multiple studies improved state-of-the-are methods, provided a better understanding of their strengths and weaknesses [6,7,9,18,22,30,45], and highlighted their applications in humanities and forensics [48,63,64]. This paper presents a review of recent advances in this field, focusing on the evaluation results of PAN shared tasks.…”

Using Learning Analytics for Preserving Academic Integrity

“…Another distinction is that the labels of the test set are also known, because student-produced content is not anonymous. The analyses were For more accurate examinations of the approach, experimental evaluation was conducted using the 10-fold cross-validation method, and used the same train/test split ratio as many of the other authorship studies (Brocardo, Traore, & Woungang, 2015;Schmid, Iqbal, & Fung, 2015). Data were randomly split into training and testing sets where 90% of the data were allocated for training and the remaining 10% allocated for testing.…”

Authorship verification using deep belief network systems

“…In our proposed n ‐gram model, we measure the degree of similarity between a block b of characters and the profile of a user U , for details see our previous work in the work of Brocardo et al We analyze whether or not a specific n ‐gram is present and compute new features by defining 2 similarity metrics: r U ( b , m ) the real‐valued similarity and d U ( b ) the binary similarity. For determining the n ‐gram, 2 modes have been considered: the unique n ‐grams mode denoted by m = 0 and the all n ‐grams mode denoted by m = 1, where m is a binary variable.…”