Adversarial stylometry

A Scientometric Study of the Stylometric Research Field

Stylometry has gained great popularity in digital humanities and social sciences. Many works on stylometry have recently been reported. However, there is a research gap regarding review studies in this field from a bibliometric and evolutionary perspective. Therefore, in this paper, a bibliometric analysis of publications from the Scopus database in the stylometric research field was proposed. Then, research articles published between 1968 and 2021 were collected and analyzed using the Bibliometrix R package for bibliometric analysis via the Biblioshiny web interface. Empirical results were also presented in terms of the performance analysis and the science mapping analysis. From these results, it is concluded that there has been a strong growth in stylometry research in recent years, while the USA, Poland, and the UK are the most productive countries, and this is due to many strong research partnerships. It was also concluded that the research topics of most articles, based on author keywords, focused on two broad thematic categories: (1) the main tasks in stylometry and (2) methodological approaches (statistics and machine learning methods).

Post-Authorship Attribution Using Regularized Deep Neural Network

Post-authorship attribution is a scientific process of using stylometric features to identify the genuine writer of an online text snippet such as an email, blog, forum post, or chat log. It has useful applications in manifold domains, for instance, in a verification process to proactively detect misogynistic, misandrist, xenophobic, and abusive posts on the internet or social networks. The process assumes that texts can be characterized by sequences of words that agglutinate the functional and content lyrics of a writer. However, defining an appropriate characterization of text to capture the unique writing style of an author is a complex endeavor in the discipline of computational linguistics. Moreover, posts are typically short texts with obfuscating vocabularies that might impact the accuracy of authorship attribution. The vocabularies include idioms, onomatopoeias, homophones, phonemes, synonyms, acronyms, anaphora, and polysemy. The method of the regularized deep neural network (RDNN) is introduced in this paper to circumvent the intrinsic challenges of post-authorship attribution. It is based on a convolutional neural network, bidirectional long short-term memory encoder, and distributed highway network. The neural network was used to extract lexical stylometric features that are fed into the bidirectional encoder to extract a syntactic feature-vector representation. The feature vector was then supplied as input to the distributed high networks for regularization to minimize the network-generalization error. The regularized feature vector was ultimately passed to the bidirectional decoder to learn the writing style of an author. The feature-classification layer consists of a fully connected network and a SoftMax function to make the prediction. The RDNN method was tested against thirteen state-of-the-art methods using four benchmark experimental datasets to validate its performance. Experimental results have demonstrated the effectiveness of the method when compared to the existing state-of-the-art methods on three datasets while producing comparable results on one dataset.

Privacy Issues in Stylometric Methods

Stylometry is a well-known field, aiming to identify the author of a text, based only on the way she/he writes. Despite its obvious advantages in several areas, such as in historical research or for copyright purposes, it may also yield privacy and personal data protection issues if it is used in specific contexts, without the users being aware of it. It is, therefore, of importance to assess the potential use of stylometry methods, as well as the implications of their use for online privacy protection. This paper aims to present, through relevant experiments, the possibility of the automated identification of a person using stylometry. The ultimate goal is to analyse the risks regarding privacy and personal data protection stemming from the use of stylometric techniques to evaluate the effectiveness of a specific stylometric identification system, as well as to examine whether proper anonymisation techniques can be applied so as to ensure that the identity of an author of a text (e.g., a user in an anonymous social network) remains hidden, even if stylometric methods are to be applied for possible re-identification.

Large-scale and Robust Code Authorship Identification with Deep Feature Learning

Successful software authorship de-anonymization has both software forensics applications and privacy implications. However, the process requires an efficient extraction of authorship attributes. The extraction of such attributes is very challenging, due to various software code formats from executable binaries with different toolchain provenance to source code with different programming languages. Moreover, the quality of attributes is bounded by the availability of software samples to a certain number of samples per author and a specific size for software samples. To this end, this work proposes a deep Learning-based approach for software authorship attribution, that facilitates large-scale, format-independent, language-oblivious, and obfuscation-resilient software authorship identification. This proposed approach incorporates the process of learning deep authorship attribution using a recurrent neural network, and ensemble random forest classifier for scalability to de-anonymize programmers. Comprehensive experiments are conducted to evaluate the proposed approach over the entire Google Code Jam (GCJ) dataset across all years (from 2008 to 2016) and over real-world code samples from 1,987 public repositories on GitHub. The results of our work show high accuracy despite requiring a smaller number of samples per author. Experimenting with source-code, our approach allows us to identify 8,903 GCJ authors, the largest-scale dataset used by far, with an accuracy of 92.3%. Using the real-world dataset, we achieved an identification accuracy of 94.38% for 745 C programmers on GitHub. Moreover, the proposed approach is resilient to language-specifics, and thus it can identify authors of four programming languages (e.g., C, C++, Java, and Python), and authors writing in mixed languages (e.g., Java/C++, Python/C++). Finally, our system is resistant to sophisticated obfuscation (e.g., using C Tigress) with an accuracy of 93.42% for a set of 120 authors. Experimenting with executable binaries, our approach achieves 95.74% for identifying 1,500 programmers of software binaries. Similar results were obtained when software binaries are generated with different compilation options, optimization levels, and removing of symbol information. Moreover, our approach achieves 93.86% for identifying 1,500 programmers of obfuscated binaries using all features adopted in Obfuscator-LLVM tool.

An Ensemble of Ensembles Approach to Author Attribution for Internet Relay Chat Forensics

With the advances in Internet technologies and services, social media has been gained extreme popularity, especially because these technologies provide potential anonymity, which in turn harbors hacker discussion forums, underground markets, dark web, and so on. Internet relay chat (IRC) is a real-time communication protocol actively used by cybercriminals for hacking, cracking, and carding. Hence, it is particularly urgent to identify the authors of threat messages and malicious activities in IRC. Unfortunately, author identification studies in IRC remain as an underexplored area. In this research, we perform novel IRC text feature extraction methods and propose the first author attribution version of the deep forest (DF) model that is an ensemble of ensembles that utilizes the fusion of ensemble learning techniques. Our approach is supported by autonomic IRC monitoring. Experiments show that our approach is highly effective for author attribution and attains high accuracy even when the number of candidates is large while training data is limited.

Identifiability, Risk, and Information Credibility in Discussions on Moral/Ethical Violation Topics on Chinese Social Networking Sites

One heated argument in recent years concerns whether requiring real name supervision on social media will inhibit users’ participation in discoursing online speech. The current study explores the impact of identification, perceived anonymity, perceived risk, and information credibility on participating in discussions on moral/ethical violation events on social network sites (SNS) in China. In this study, we constructed a model based on the literature and tested it on a sample of 218 frequent SNS users. The results demonstrate the influence of identification and perception of anonymity: although the relationship between the two factors is negative, both are conducive to participation in discussion on moral/ethical violation topics, and information credibility also has a positive impact. The results confirmed the significance of risk perception on comments posted about moral/ethical violation. Our results have reference value for identity management and internet governance. Policies regarding users’ real names on the internet need to take into account the reliability of the identity authentication mechanism, as well as netizens’ perceptions of privacy about their identity and the necessity of guaranteeing content and information reliability online. We also offer some suggestions for future research, with a special emphasis on applicability to different cultures, contexts, and social networking sites.

Effective writing style transfer via combinatorial paraphrasing

Stylometry can be used to profile or deanonymize authors against their will based on writing style. Style transfer provides a defence. Current techniques typically use either encoder-decoder architectures or rule-based algorithms. Crucially, style transfer must reliably retain original semantic content to be actually deployable. We conduct a multifaceted evaluation of three state-of-the-art encoder-decoder style transfer techniques, and show that all fail at semantic retainment. In particular, they do not produce appropriate paraphrases, but only retain original content in the trivial case of exactly reproducing the text. To mitigate this problem we propose ParChoice: a technique based on the combinatorial application of multiple paraphrasing algorithms. ParChoice strongly outperforms the encoder-decoder baselines in semantic retainment. Additionally, compared to baselines that achieve nonnegligible semantic retainment, ParChoice has superior style transfer performance. We also apply ParChoice to multi-author style imitation (not considered by prior work), where we achieve up to 75% imitation success among five authors. Furthermore, when compared to two state-of-the-art rule-based style transfer techniques, ParChoice has markedly better semantic retainment. Combining ParChoice with the best performing rulebased baseline (Mutant-X [34]) also reaches the highest style transfer success on the Brennan-Greenstadt and Extended-Brennan-Greenstadt corpora, with much less impact on original meaning than when using the rulebased baseline techniques alone. Finally, we highlight a critical problem that afflicts all current style transfer techniques: the adversary can use the same technique for thwarting style transfer via adversarial training. We show that adding randomness to style transfer helps to mitigate the effectiveness of adversarial training.

Multi-χ: Identifying Multiple Authors from Source Code Files

Most authorship identification schemes assume that code samples are written by a single author. However, real software projects are typically the result of a team effort, making it essential to consider a finegrained multi-author identification in a single code sample, which we address with Multi-χ. Multi-χ leverages a deep learning-based approach for multi-author identification in source code, is lightweight, uses a compact representation for efficiency, and does not require any code parsing, syntax tree extraction, nor feature selection. In Multi-χ, code samples are divided into small segments, which are then represented as a sequence of n-dimensional term representations. The sequence is fed into an RNN-based verification model to assist a segment integration process which integrates positively verified segments, i.e., integrates segments that have a high probability of being written by one author. Finally, the resulting segments from the integration process are represented using word2vec or TF-IDF and fed into the identification model. We evaluate Multi-χ with several Github projects (Caffe, Facebook’s Folly, Tensor-Flow, etc.) and show remarkable accuracy. For example, Multi-χ achieves an authorship example-based accuracy (A-EBA) of 86.41% and per-segment authorship identification of 93.18% for identifying 562 programmers. We examine the performance against multiple dimensions and design choices, and demonstrate its effectiveness.

On divergence-based author obfuscation: An attack on the state of the art in statistical authorship verification

Authorship verification is the task of determining whether two texts were written by the same author based on a writing style analysis. Author obfuscation is the adversarial task of preventing a successful verification by altering a text’s style so that it does not resemble that of its original author anymore. This paper introduces new algorithms for both tasks and reports on a comprehensive evaluation to ascertain the merits of the state of the art in authorship verification to withstand obfuscation.

After introducing a new generalization of the well-known unmasking algorithm for short texts, thus completing our collection of state-of-the-art algorithms for verification, we introduce an approach that (1) models writing style difference as the Jensen-Shannon distance between the character n-gram distributions of texts, and (2) manipulates an author’s writing style in a sophisticated manner using heuristic search. For obfuscation, we explore the huge space of textual variants in order to find a paraphrased version of the to-be-obfuscated text that has a sufficiently high Jensen-Shannon distance at minimal costs in terms of text quality loss. We analyze, quantify, and illustrate the rationale of this approach, define paraphrasing operators, derive text length-invariant thresholds for termination, and develop an effective obfuscation framework. Our authorship obfuscation approach defeats the presented state-of-the-art verification approaches, while keeping text changes at a minimum. As a final contribution, we discuss and experimentally evaluate a reverse obfuscation attack against our obfuscation approach as well as possible remedies.

A Girl Has No Name: Automated Authorship Obfuscation using Mutant-X

Stylometric authorship attribution aims to identify an anonymous or disputed document’s author by examining its writing style. The development of powerful machine learning based stylometric authorship attribution methods presents a serious privacy threat for individuals such as journalists and activists who wish to publish anonymously. Researchers have proposed several authorship obfuscation approaches that try to make appropriate changes (e.g. word/phrase replacements) to evade attribution while preserving semantics. Unfortunately, existing authorship obfuscation approaches are lacking because they either require some manual effort, require significant training data, or do not work for long documents. To address these limitations, we propose a genetic algorithm based random search framework called Mutant-X which can automatically obfuscate text to successfully evade attribution while keeping the semantics of the obfuscated text similar to the original text. Specifically, Mutant-X sequentially makes changes in the text using mutation and crossover techniques while being guided by a fitness function that takes into account both attribution probability and semantic relevance. While Mutant-X requires black-box knowledge of the adversary’s classifier, it does not require any additional training data and also works on documents of any length. We evaluate Mutant-X against a variety of authorship attribution methods on two different text corpora. Our results show that Mutant-X can decrease the accuracy of state-of-the-art authorship attribution methods by as much as 64% while preserving the semantics much better than existing automated authorship obfuscation approaches. While Mutant-X advances the state-of-the-art in automated authorship obfuscation, we find that it does not generalize to a stronger threat model where the adversary uses a different attribution classifier than what Mutant-X assumes. Our findings warrant the need for future research to improve the generalizability (or transferability) of automated authorship obfuscation approaches.

Does Identification Influence Continuous E-Commerce Consumption? The Mediating Role of Intrinsic Motivations

The motivation behind online consumption behavior is different from that of online social behavior, and research is lacking regarding the impact of identification on e-commerce consumption. The current research examines the influence of identification, which is perceived anonymity, and intrinsic motivation on the continuous purchasing behaviors on retailing e-commerce websites based on self-determination theory. The mediating role of intrinsic motivation was also empirically tested from a sample of 661 frequent consumers using the partial least squares approach. The findings were: (1) Identification negatively influences perceived anonymity, and its low, but significantly positive, influence on continuous e-commerce consumption were totally mediated by perceived competence, perceived autonomy, and perceived relatedness. (2) Perceived anonymity positively influences self-determination factors, which has partly mediating impact between perceived anonymity and continuous consumption. (3) The authenticity and concealment of identity are based on different mechanisms, but both of them are conducive to promoting continuous purchases. On retailing e-commerce websites, customers’ identity management should consider both identification in the background and anonymity perception in the service, and the contributions of the service to promote consumers’ perceived competence and perceived autonomy are important in continuous consumption.

Network motifs for translator stylometry identification

Despite the extensive literature investigating stylometry analysis in authorship attribution research, translator stylometry is an understudied research area. The identification of translator stylometry contributes to many fields including education, intellectual property rights and forensic linguistics. In a two stage process, this paper first evaluates the use of existing lexical measures for the translator stylometry problem. Similar to previous research we found that using vocabulary richness in its traditional form as it has been used in the literature could not identify translator stylometry. This encouraged us to design an approach with the aim of identifying the distinctive patterns of a translator by employing network-motifs. Networks motifs are small sub-graphs which aim at capturing the local structure of a complex network. The proposed approach achieved an average accuracy of 83% in three-way classification. These results demonstrate that classic tools based on lexical features can be used for identifying translator stylometry if they get augmented with appropriate non-parametric scaling. Moreover, the use of complex network analysis and network motifs mining provided made it possible to design features that can solve translator stylometry analysis problems.

Learning Stylometric Representations for Authorship Analysis

Authorship analysis (AA) is the study of unveiling the hidden properties of authors from textual data. It extracts an author's identity and sociolinguistic characteristics based on the reflected writing styles in the text. The process is essential for various areas, such as cybercrime investigation, psycholinguistics, political socialization, etc. However, most of the previous techniques critically depend on the manual feature engineering process. Consequently, the choice of feature set has been shown to be scenario- or dataset-dependent. In this paper, to mimic the human sentence composition process using a neural network approach, we propose to incorporate different categories of linguistic features into distributed representation of words in order to learn simultaneously the writing style representations based on unlabeled texts for AA. In particular, the proposed models allow topical, lexical, syntactical, and character-level feature vectors of each document to be extracted as stylometrics. We evaluate the performance of our approach on the problems of authorship characterization, authorship identification and authorship verification with the Twitter, blog, review, novel, and essay datasets. The experiments suggest that our proposed text representation outperforms the static stylometrics, dynamic n -grams, latent Dirichlet allocation, latent semantic analysis, distributed memory model of paragraph vectors, distributed bag of words version of paragraph vector, word2vec representations, and other baselines.

First Steps towards Data-Driven Adversarial Deduplication

In traditional databases, the entity resolution problem (which is also known as deduplication) refers to the task of mapping multiple manifestations of virtual objects to their corresponding real-world entities. When addressing this problem, in both theory and practice, it is widely assumed that such sets of virtual objects appear as the result of clerical errors, transliterations, missing or updated attributes, abbreviations, and so forth. In this paper, we address this problem under the assumption that this situation is caused by malicious actors operating in domains in which they do not wish to be identified, such as hacker forums and markets in which the participants are motivated to remain semi-anonymous (though they wish to keep their true identities secret, they find it useful for customers to identify their products and services). We are therefore in the presence of a different, and even more challenging, problem that we refer to as adversarial deduplication. In this paper, we study this problem via examples that arise from real-world data on malicious hacker forums and markets arising from collaborations with a cyber threat intelligence company focusing on understanding this kind of behavior. We argue that it is very difficult—if not impossible—to find ground truth data on which to build solutions to this problem, and develop a set of preliminary experiments based on training machine learning classifiers that leverage text analysis to detect potential cases of duplicate entities. Our results are encouraging as a first step towards building tools that human analysts can use to enhance their capabilities towards fighting cyber threats.

Authorship attribution based on Life-Like Network Automata

The authorship attribution is a problem of considerable practical and technical interest. Several methods have been designed to infer the authorship of disputed documents in multiple contexts. While traditional statistical methods based solely on word counts and related measurements have provided a simple, yet effective solution in particular cases; they are prone to manipulation. Recently, texts have been successfully modeled as networks, where words are represented by nodes linked according to textual similarity measurements. Such models are useful to identify informative topological patterns for the authorship recognition task. However, there is no consensus on which measurements should be used. Thus, we proposed a novel method to characterize text networks, by considering both topological and dynamical aspects of networks. Using concepts and methods from cellular automata theory, we devised a strategy to grasp informative spatio-temporal patterns from this model. Our experiments revealed an outperformance over structural analysis relying only on topological measurements, such as clustering coefficient, betweenness and shortest paths. The optimized results obtained here pave the way for a better characterization of textual networks.

Recognizing and Imitating Programmer Style: Adversaries in Program Authorship Attribution

Source code attribution classifiers have recently become powerful. We consider the possibility that an adversary could craft code with the intention of causing a misclassification, i.e., creating a forgery of another author’s programming style in order to hide the forger’s own identity or blame the other author. We find that it is possible for a non-expert adversary to defeat such a system. In order to inform the design of adversarially resistant source code attribution classifiers, we conduct two studies with C/C++ programmers to explore the potential tactics and capabilities both of such adversaries and, conversely, of human analysts doing source code authorship attribution. Through the quantitative and qualitative analysis of these studies, we (1) evaluate a state-of-the-art machine classifier against forgeries, (2) evaluate programmers as human analysts/forgery detectors, and (3) compile a set of modifications made to create forgeries. Based on our analyses, we then suggest features that future source code attribution systems might incorporate in order to be adversarially resistant.