The DARPA Twitter Bot Challenge

Lecture Notes in Computer Science, 2016

The popularity of social media platforms such as Twitter has led to the proliferation of automated bots, creating both opportunities and challenges in information dissemination, user engagements, and quality of services. Past works on profiling bots had been focused largely on malicious bots, with the assumption that these bots should be removed. In this work, however, we find many bots that are benign, and propose a new, broader categorization of bots based on their behaviors. This includes broadcast, consumption, and spam bots. To facilitate comprehensive analyses of bots and how they compare to human accounts, we develop a systematic profiling framework that includes a rich set of features and classifier bank. We conduct extensive experiments to evaluate the performances of different classifiers under varying time windows, identify the key features of bots, and infer about bots in a larger Twitter population. Our analysis encompasses more than 159K bot and human (non-bot) accounts in Twitter. The results provide interesting insights on the behavioral traits of both benign and malicious bots.

PLOS ONE, 2017

It has recently become possible to study the dynamics of information diffusion in techno-social systems at scale, due to the emergence of online platforms, such as Twitter, with millions of users. One question that systematically recurs is whether information spreads according to simple or complex dynamics: does each exposure to a piece of information have an independent probability of a user adopting it (simple contagion), or does this probability depend instead on the number of sources of exposure, increasing above some threshold (complex contagion)? Most studies to date are observational and, therefore, unable to disentangle the effects of confounding factors such as social reinforcement, homophily, limited attention, or network community structure. Here we describe a novel controlled experiment that we performed on Twitter using 'social bots' deployed to carry out coordinated attempts at spreading information. We propose two Bayesian statistical models describing simple and complex contagion dynamics, and test the competing hypotheses. We provide experimental evidence that the complex contagion model describes the observed information diffusion behavior more accurately than simple contagion. Future applications of our results include more effective defenses against malicious propaganda campaigns on social media, improved marketing and advertisement strategies, and design of effective network intervention techniques.

2021

Social networks are real-time platforms formed by users involving conversations and interactions. ,is phenomenon of the new information era results in a very huge amount of data in different forms andmodalities such as text, images, videos, and voice.,e data with such characteristics are also known as big data with 5-V properties and in some cases are also referred to as social big data. To find useful information from such valuable data, many researchers tried to address different aspects of it for different modalities. In the case of text, NLP researchers conducted many research studies and scientific works to extract valuable information such as topics. Many enlightening works on different platforms of social media, like Twitter, tried to address the problem of finding important topics from different aspects and utilized it to propose solutions for diverse use cases. ,e importance of Twitter in this scope lies in its content and the behavior of its users. For example, it is also ...

International Journal of Environmental Research and Public Health

During the COVID-19 pandemic, when individuals were confronted with social distancing, social media served as a significant platform for expressing feelings and seeking emotional support. However, a group of automated actors known as social bots have been found to coexist with human users in discussions regarding the coronavirus crisis, which may pose threats to public health. To figure out how these actors distorted public opinion and sentiment expressions in the outbreak, this study selected three critical timepoints in the development of the pandemic and conducted a topic-based sentiment analysis for bot-generated and human-generated tweets. The findings show that suspected social bots contributed to as much as 9.27% of COVID-19 discussions on Twitter. Social bots and humans shared a similar trend on sentiment polarity—positive or negative—for almost all topics. For the most negative topics, social bots were even more negative than humans. Their sentiment expressions were weaker ...

Advanced Information Systems Engineering

Chatbot applications are increasingly adopted in various domains such as e-commerce or customer services as a direct communication channel between companies and end-users. Multiple frameworks have been developed to ease their definition and deployment. They typically rely on existing cloud infrastructures and artificial intelligence techniques to efficiently process user inputs and extract conversation information. While these frameworks are efficient to design simple chatbot applications, they still require advanced technical knowledge to define complex conversations and interactions. In addition, the deployment of a chatbot application usually requires a deep understanding of the targeted platforms, increasing the development and maintenance costs. In this paper we introduce the Jarvis framework, that tackles these issues by providing a Domain Specific Language (DSL) to define chatbots in a platform-independent way, and a runtime engine that automatically deploys the chatbot application and manages the defined conversation logic. Jarvis is open source and fully available online.

Informatics

Bots (social robots) are computer programs that replicate human behavior in online social networks. They are either fully automated or semi-automated, and their use makes online activism vulnerable to manipulation. This study examines the existence of social robots in the #FeesMustFall movement by conducting a scientific investigation into whether social bots were present in the form of Twitter bots and cyborgs. A total of 576,823 tweets posted between 15 October 2015 and 10 April 2017 were cleaned, with 490,449 tweets analyzed for 90,783 unique persons. Three separate approaches were used to screen out suspicious bot and cyborg activity, supplemented by the DeBot team’s methodology. User 1 and User 2, two of the 90,783 individuals, were recognized as bots or cyborgs in the study and contributed 22,413 (4.57 percent) of the 490,449 tweets. This confirms the existence of bots throughout the campaign, which aided in the #FeesMustFall’s amplification on Twitter, complicating sentiment ...

International Journal of Disaster Risk Reduction, 2021

Streaming social media provides a real-time glimpse of extreme weather impacts. However, the volume of streaming data makes mining information a challenge for emergency managers, policy makers, and disciplinary scientists. Here we explore the effectiveness of data learned approaches to mine and filter information from streaming social media data from Hurricane Irma's landfall in Florida, USA. We use 54,383 Twitter messages (out of 784 K geolocated messages) from 16,598 users from Sept. 10-12, 2017 to develop 4 independent models to filter data for relevance: 1) a geospatial model based on forcing conditions at the place and time of each tweet, 2) an image classification model for tweets that include images, 3) a user model to predict the reliability of the tweeter, and 4) a text model to determine if the text is related to Hurricane Irma. All four models are independently tested, and can be combined to quickly filter and visualize tweets based on user-defined thresholds for each submodel. We envision that this type of filtering and visualization routine can be useful as a base model for data capture from noisy sources such as Twitter. The data can then be subsequently used by policy makers, environmental managers, emergency managers, and domain scientists interested in finding tweets with specific attributes to use during different stages of the disaster (e.g., preparedness, response, and recovery), or for detailed research.

Social Network Analysis and Mining, 2019

Until recently, social media were seen to promote democratic discourse on social and political issues. However, this powerful communication ecosystem has come under scrutiny for allowing hostile actors to exploit online discussions in an attempt to manipulate public opinion. A case in point is the ongoing U.S. Congress investigation of Russian interference in the 2016 U.S. election campaign, with Russia accused of, among other things, using trolls (malicious accounts created for the purpose of manipulation) and bots (automated accounts) to spread propaganda and politically biased information. In this study, we explore the effects of this manipulation campaign, taking a closer look at users who re-shared the posts produced on Twitter by the Russian troll accounts publicly disclosed by U.S. Congress investigation. We collected a dataset of 13 million election-related posts shared on Twitter in the year of 2016 by over a million distinct users. This dataset includes accounts associated with the identified Russian trolls as well as users sharing posts in the same time period on a variety of topics around the 2016 elections. We use label propagation to infer the users' ideology based on the news sources they share. We are able to classify a large number of the users as liberal or con

Lecture Notes in Computer Science, 2016

We present a machine learning framework that leverages a mixture of metadata, network, and temporal features to detect extremist users, and predict content adopters and interaction reciprocity in social media. We exploit a unique dataset containing millions of tweets generated by more than 25 thousand users who have been manually identified, reported, and suspended by Twitter due to their involvement with extremist campaigns. We also leverage millions of tweets generated by a random sample of 25 thousand regular users who were exposed to, or consumed, extremist content. We carry out three forecasting tasks, (i) to detect extremist users, (ii) to estimate whether regular users will adopt extremist content, and finally (iii) to predict whether users will reciprocate contacts initiated by extremists. All forecasting tasks are set up in two scenarios: a post hoc (time independent) prediction task on aggregated data, and a simulated real-time prediction task. The performance of our framework is extremely promising, yielding in the different forecasting scenarios up to 93% AUC for extremist user detection, up to 80% AUC for content adoption prediction, and finally up to 72% AUC for interaction reciprocity forecasting. We conclude by providing a thorough feature analysis that helps determine which are the emerging signals that provide predictive power in different scenarios.

SN Computer Science

A very large number of people use Online Social Networks daily. Such platforms thus become attractive targets for agents that seek to gain access to the attention of large audiences, and influence perceptions or opinions. Botnets, collections of automated accounts controlled by a single agent, are a common mechanism for exerting maximum influence. Botnets may be used to better infiltrate the social graph over time and to create an illusion of community behavior, amplifying their message and increasing persuasion. This paper investigates Twitter botnets, their behavior, their interaction with user communities and their evolution over time. We analyzed a dense crawl of a subset of Twitter traffic, amounting to nearly all interactions by Greek-speaking Twitter users for a period of 36 months. We detected over a million events where seemingly unrelated accounts tweeted nearly identical content at nearly the same time. We filtered these concurrent content injection events and detected a set of 1,850 accounts that repeatedly exhibit this pattern of behavior, suggesting that they are fully or in part controlled and orchestrated by the same software. We found botnets that appear for brief intervals and disappear, as well as botnets that evolve and grow, spanning the duration of our dataset. We analyze statistical differences between bot accounts and human users, as well as botnet interaction with user communities and Twitter trending topics.

Comunicar, 2021

Twitter has transformed into one of the main platforms for citizen engagement today. However, even though previous studies have focused on opinions about vaccines in general or about specific vaccines, opinions towards COVID-19 vaccines on Twitter have not been researched to date. The objective of this research is, by using social network analysis and language processing tools, to examine the degree to which the opinions and interactions present on Twitter are favorable or unfavorable towards the main COVID-19 vaccines. In addition, the relevance of each of the vaccines is studied, as well as their level of controversy. Likewise, the present study investigates, for the first time, the conversation from different perspectives including the content and also the participants, by analyzing in detail the verified accounts and using tools for the detection of bots. In global terms, the results from verified accounts show a moderate favorability towards the COVID-19 vaccines, the most acce...

Big Data and Cognitive Computing, 2022

Social media platforms are part of everyday life, allowing the interconnection of people around the world in large discussion groups relating to every topic, including important social or political issues. Therefore, social media have become a valuable source of information-rich data, commonly referred to as Social Big Data, effectively exploitable to study the behavior of people, their opinions, moods, interests and activities. However, these powerful communication platforms can be also used to manipulate conversation, polluting online content and altering the popularity of users, through spamming activities and misinformation spreading. Recent studies have shown the use on social media of automatic entities, defined as social bots, that appear as legitimate users by imitating human behavior aimed at influencing discussions of any kind, including political issues. In this paper we present a new methodology, namely TIMBRE (Time-aware opInion Mining via Bot REmoval), aimed at discove...

Electronics, 2022

In the ultra-connected age of information, online social media platforms have become an indispensable part of our daily routines. Recently, this online public space is getting largely occupied by suspicious and manipulative social media bots. Such automated deceptive bots often attempt to distort ground realities and manipulate global trends, thus creating astroturfing attacks on the social media online portals. Moreover, these bots often tend to participate in duplicitous activities, including promotion of hidden agendas and indulgence in biased propagation meant for personal gain or scams. Thus, online bots have eventually become one of the biggest menaces for social media platforms. Therefore, we have proposed an AI-driven social media bot identification framework, namely TweezBot, which can identify fraudulent Twitter bots. The proposed bot detection method analyzes Twitter-specific user profiles having essential profile-centric features and several activity-centric characterist...

British Journal of Management, 2021

This paper has been accepted at the British Journal of Management. It is on the production stage and will go online soon.

Information Systems Frontiers, 2020

Malicious behavior in social media has many faces, which for instance appear in the form of bots, sock puppets, creation and dissemination of fake news, Sybil attacks, and actors hiding behind multiple identities. In this paper, we propose the NETDER architecture, which takes its name from its two main modules: Network Diffusion and ontological reasoning based on Existential Rules), to address these issues. This initial proposal is meant to serve as a roadmap for research and development of tools to attack malicious behavior in social media, guiding the implementation of software in this domain, instead of a specific solution. Our working hypothesis is that these problems-and many others-can be effectively tackled by (i) combining multiple data sources that are constantly being updated, (ii) maintaining a knowledge base using logicbased formalisms capable of value invention to support generating hypotheses based on available data, and (iii) maintaining a related knowledge base with information regarding how actors are connected, and how information flows across their network. We show how these three basic tenets give rise to a general model that has the further capability of addressing multiple problems at once.

Journal of Medical Internet Research, 2021

As of March 2021, the SARS-CoV-2 virus has been responsible for over 115 million cases of COVID-19 worldwide, resulting in over 2.5 million deaths. As the virus spread exponentially, so did its media coverage, resulting in a proliferation of conflicting information on social media platforms-a so-called "infodemic." In this viewpoint, we survey past literature investigating the role of automated accounts, or "bots," in spreading such misinformation, drawing connections to the COVID-19 pandemic. We also review strategies used by bots to spread (mis)information and examine the potential origins of bots. We conclude by conducting and presenting a secondary analysis of data sets of known bots in which we find that up to 66% of bots are discussing COVID-19. The proliferation of COVID-19 (mis)information by bots, coupled with human susceptibility to believing and sharing misinformation, may well impact the course of the pandemic.

International Journal of Community Well-Being, 2020

This paper examines the role and potential risk of artificial intelligence (AI) powered automated social media accounts in participatory planning processes and broader themes of community well-being. The rapid growth and massive uptake of social media has resulted in a surge in public interest to engage with others around key land development topics. Due to the low cost and high potential engagement, planners and policymakers have been quick to open electronic channels of participation to inform the decision-making process. Doing so has created an opportunity for subversion from groups with alternate and possibly nefarious interests. Anecdotally, we have found that automated social media accounts have been used to further inflate the voice, and therefore influence, of subversive groups in the land development and planning process. While scholars have begun to examine how tech-savvy social media users are manipulating political discourse through the medium of Facebook and Twitter, a dearth of research has yet sought to examine the potential harm that such manipulation could cause to online planning processes and resulting community well-being. This project seeks to explore the risks that social media manipulation might pose to online community discourse around land development and planning topics. To begin to gauge this risk, this paper reviews the theoretical and empirical literature on the topic and makes recommendations for future research to measure and analyze the threat to Twitter community well-being posed by AI.

Social Network Analysis and Mining

In today’s digitalized era, Online Social Networking platforms are growing to be a vital aspect of each individual’s daily life. The availability of the vast amount of information and their open nature attracts the interest of cybercriminals to create malicious bots. Malicious bots in these platforms are automated or semi-automated entities used in nefarious ways while simulating human behavior. Moreover, such bots pose serious cyber threats and security concerns to society and public opinion. They are used to exploit vulnerabilities for illicit benefits such as spamming, fake profiles, spreading inappropriate/false content, click farming, hashtag hijacking, and much more. Cybercriminals and researchers are always engaged in an arms race as new and updated bots are created to thwart ever-evolving detection technologies. This literature review attempts to compile and compare the most recent advancements in Machine Learning-based techniques for the detection and classification of bots...

Nature Communications, 2018

The massive spread of digital misinformation has been identified as a major threat to democracies. Communication, cognitive, social, and computer scientists are studying the complex causes for the viral diffusion of misinformation, while online platforms are beginning to deploy countermeasures. Little systematic, data-based evidence has been published to guide these efforts. Here we analyze 14 million messages spreading 400 thousand articles on Twitter during ten months in 2016 and 2017. We find evidence that social bots played a disproportionate role in spreading articles from low-credibility sources. Bots amplify such content in the early spreading moments, before an article goes viral. They also target users with many followers through replies and mentions. Humans are vulnerable to this manipulation, resharing content posted by bots. Successful low-credibility sources are heavily supported by social bots. These results suggest that curbing social bots may be an effective strategy...

Human Behavior and Emerging Technologies, 2019

Proceedings of the 29th ACM International Conference on Information & Knowledge Management, 2020

Malicious actors create inauthentic social media accounts controlled in part by algorithms, known as social bots, to disseminate misinformation and agitate online discussion. While researchers have developed sophisticated methods to detect abuse, novel bots with diverse behaviors evade detection. We show that different types of bots are characterized by different behavioral features. As a result, supervised learning techniques suffer severe performance deterioration when attempting to detect behaviors not observed in the training data. Moreover, tuning these models to recognize novel bots requires retraining with a significant amount of new annotations, which are expensive to obtain. To address these issues, we propose a new supervised learning method that trains classifiers specialized for each class of bots and combines their decisions through the maximum rule. The ensemble of specialized classifiers (ESC) can better generalize, leading to an average improvement of 56% in F1 score for unseen accounts across datasets. Furthermore, novel bot behaviors are learned with fewer labeled examples during retraining. We deployed ESC in the newest version of Botometer, a popular tool to detect social bots in the wild, with a cross-validation AUC of 0.99.

Applied Sciences, 2021

This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY

Studies in Computational Intelligence, 2018

The emergence of social bots within online social networks (OSNs) to diffuse information at scale has given rise to many efforts to detect them. While methodologies employed to detect the evolving sophistication of bots continue to improve, much work can be done to characterize the impact of bots on communication networks. In this study, we present a framework to describe the pervasiveness and relative importance of participants recognized as bots in various OSN conversations. Specifically, we harvested over 30 million tweets from three major global events in 2016 (the U.S. Presidential Election, the Ukrainian Conflict and Turkish Political Censorship) and compared the conversational patterns of bots and humans within each event. We further examined the social network structure of each conversation to determine if bots exhibited any particular network influence, while also determining bot participation in key emergent network communities. The results showed that although participants recognized as social bots comprised only 0.28% of all OSN users in this study, they accounted for a significantly large portion of prominent centrality rankings across the three conversations. This includes the identification of individual bots as top-10 influencer nodes out of a total corpus consisting of more than 2.8 million nodes.

Social Network Analysis and Mining, 2021

Due to their popularity, online social networks are a popular target for spam, scams, malware distribution and more recently state-actor propaganda. In this paper, we review a number of recent approaches to fake account and bot classification. Based on this review and our experiments, we propose our own method which leverages the social graph's topology and differences in egographs of legitimate and fake user accounts to improve identification of the latter. We evaluate our approach against other common approaches on a real-world dataset of users of the social network Twitter.

Applied Intelligence, 2019

In an online social network (like Twitter), a botmaster (i.e., leader among a group of social bots) establishes a social relationship among legitimate participants to reduce the probability of social bot detection. Social bots generate fake tweets and spread malicious information by manipulating the public opinion. Therefore, the detection of social bots in an online social network is an important task. In this paper, we consider social attributes, such as tweet-based attributes, user profilebased attributes and social graph-based attributes for detecting the social bots among legitimate participants. We design a deep Q-network architecture by incorporating a Deep Q-Learning (DQL) model using the social attributes in the Twitter network for detection of social bots based on updating Q-value function (i.e., state-action value function). We consider each social attribute of a user as a state and the learning agent's movement from one state to another state is considered as an action. For Q-value function, we consider all the state-action pairs in order to construct the state transition probability values between the state-action pairs. In the proposed DQL algorithm, the learning agent chooses a specific learning action with an optimal Q-value in each state for social bot detection. Further, we also propose an approach that identifies the most influential users (which are influenced by the social bots) based on tweets and the users' interactions. The experimentation using the datasets collected from Twitter network illustrates the efficacy of proposed model.

2020 3rd International Conference on Data Intelligence and Security (ICDIS)

Over the past few years, we have observed different media outlets' attempts to shift public opinion by framing information to support a narrative that facilitate their goals. Malicious users referred to as "pathogenic social media" (PSM) accounts are more likely to amplify this phenomena by spreading misinformation to viral proportions. Understanding the spread of misinformation from account-level perspective is thus a pressing problem. In this work, we aim to present a feature-driven approach to detect PSM accounts in social media. Inspired by the literature, we set out to assess PSMs from three broad perspectives: (1) user-related information (e.g., user activity, profile characteristics), (2) source-related information (i.e., information linked via URLs shared by users) and (3) content-related information (e.g., tweets characteristics). For the user-related information, we investigate malicious signals using causality analysis (i.e., if user is frequently a cause of viral cascades) and profile characteristics (e.g., number of followers, etc.). For the source-related information, we explore various malicious properties linked to URLs (e.g., URL address, content of the associated website, etc.). Finally, for the content-related information, we examine attributes (e.g., number of hashtags, suspicious hashtags, etc.) from tweets posted by users. Experiments on real-world Twitter data from different countries demonstrate the effectiveness of the proposed approach in identifying PSM users.

Multimodal technologies and interaction, 2020

After conversational agents have been made available to the broader public, we speculate that applying them as a mediator for adaptive environments reduces control complexity and increases user experience by providing a more natural interaction. We implemented and tested four agents, each of them differing in their system intelligence and input modality, as personal assistants for Mediated Atmospheres, an adaptive smart office prototype. They were evaluated in a user study (N = 33) to collect subjective and objective measures. Results showed that a smartphone application was the most favorable system, followed by conversational text and voice agents that were perceived as being more engaging and intelligent than a non-conversational voice agent. Significant differences were observed between native and non-native speakers in both subjective and objective measures. Our findings reveal the potential of conversational agents for the interaction with adaptive environments to reduce work and information overload.

Social Network Analysis and Mining, 2016

Online social networks (OSNs) such as Twitter and Facebook have become a significant testing ground for Artificial Intelligence developers who build programs, known as socialbots, that imitate human users by automating their social network activities such as forming social links and posting content. Particularly, Twitter users have shown difficulties in distinguishing these socialbots from the human users in their social graphs. Frequently, socialbots are effective in acquiring human users as followers and exercising influence within them. While the success of socialbots is certainly a remarkable achievement for AI practitioners, their proliferation in the Twitter sphere opens many possibilities for cybercrime. The proliferation of socialbots in Twitter motivates us to assess the characteristics or strategies that make socialbots most likely to succeed. In this direction, we created 120 socialbot accounts in Twitter, which have a profile, follow other users, and generate tweets either by reposting others' tweets or by generating their own synthetic tweets. Then, we employ a 2 k factorial design experiment to quantify the infiltration performance of different socialbot strategies, and examine the effectiveness of individual profile and activity-related attributes of the socialbots. Our analysis is the first of a kind, and reveals what strategies make socialbots successful in the Twitter sphere.