Speaker: Gianluca Stringhini, Boston University

Title: Computational Methods to Understand and Mitigate Online Hate

Abstract:Online hate on social media is becoming a serious problem. Aggression is often not the act of single individuals, but rather the result of coordinated activity between like minded people who gather on polarized online communities, identify suitable targets, and carry out their attacks. Studying this phenomenon is hard, in part because it is not a purely technical problem, in part because this malicious activity unfolds across multiple online services, and the research community currently lacks effective tools to keep track of information that is not contained to a single platform. In this talk, I will present our research on studying the modus operandi of attackers that orchestrate and execute coordinated hate attacks on social media, which we call "raids." I will show that online polarized communities such as 4chan's politically incorrect board are particularly effective in performing this type of attack, often causing great psychological harm to their victims, and that it is possible to identify raids by looking at coordinated activity happening on different platforms (e.g., a thread on 4chan and the comments on a YouTube video). Finally, I will discuss potential countermeasures against this phenomenon. (Slides Can be found here)

Bio:Gianluca Stringhini is an Assistant Professor in the Electrical and Computer Engineering Department at Boston University. Before joining BU, he was faculty at University College London. Gianluca works in the area of data-driven security, analyzing large datasets to better understand complex malicious online operations and developing mitigation techniques to fight them. He was awarded a Facebook Secure the Internet Grant in 2018, a Google Faculty Research Award in 2015, the Symantec Research Labs Fellowship in 2012, and multiple Best Paper Awards, including one at IMC 2018. He has published in top security conferences such as CCS, NDSS, and USENIX Security, as well as top measurement and Web conferences such as IMC, ICWSM, and WWW.

Speaker: Barry Bradlyn, Assistant Professor, Physics Dept., University of Illinois at Urbana-Champaign

Title: Statistical Mechanics of Sh*tposting

Abstract: With the rise of social media, the speed and scale of information propagation have risen to unprecedented levels. Along with all its positive societal effects, this communication revolution has also unfortunately created platforms for extremists to congregate, recruit, and plan. Understanding the formation and growth of these extremist networks, as well as decoding the meaning of their slang and visual imagery, is of utmost priority. In this talk, I outline a transdisciplinary research program to address these issues, combining tools from both computer science and statistical physics. I will review our prior work showing how semantic embeddings and stochastic process modeling can be used to track the meaning and influence of memes across different social networks. I will then outline our efforts to incorporate time-dependence into these models, in order to create a coarse-grained, generative model of information diffusion in extremist networks.

Bio: Barry Bradlyn is an Assistant Professor in the physics department at the University of Illinois, a primary member of the International Data-driven Research for Advanced Modeling and Analysis Lab, and a founding member of the Network Contagion Research Institute. His has worked on topics ranging from developing a formalism for computing the viscoelastic and thermal response functions for two dimensional topological phases, predicting the existence of topologically charged, multiply degenerate fermions in weakly interacting crystals with no known analogue in high energy physics, and developing real-space formulation of topological band theory. More recently, his work with iDRAMA Lab and NCRI has focused on developing numerical methods to analyze large-scale, human behavior through the lens of hate speech and racism online.

Barry received his bachelor’s degree in Physics from MIT in 2009 and his PhD in physics from Yale University. From 2015 to 2018, Barry was a postdoctoral researcher at the Princeton Center for Theoretical Science.

Speaker:Tyler Otto

Title: Content Security at Reddit

Abstract:Reddit’s moderation and governance model is unique within the internet landscape. In addition to its sitewide policies and enforcement, Reddit has cultivated a community-driven ecosystem that empowers users to work with volunteer community moderators and internal company admins to help curate quality content and reduce the exposure of rule-breaking content and ensure the content is fitting with community standards. Increasingly, Reddit is using data science to aggregate and enrich these signals, allowing us to catch issues that would be more challenging for an individual to notice. This investment in data science has drastically improved the speed and efficacy with which Reddit is able to respond to threats. By better leveraging analytics, Reddit has been able to increase the percentage of content that is taken down proactively from 27% in 2017 to around 99%.

Bio:Tyler Otto is the Director of Threat Analytics at Reddit. His team leverages data science for the detection and mitigation of scaled threats targeting Reddit. This includes things like spam, vote manipulation, site integrity, and malicious bots. Additionally, his team builds models to help improve traditional operations teams be more effective and efficient. Prior to running threat analytics at Reddit, Tyler lead the product Data Science team at Reddit. Before Reddit, he was the Head of Data Science at Hipmunk, a travel startup that was acquired by Concur.

Speaker:Brad Chen

Title: YouTube Trust and Safety Engineering

Abstract: YouTube Trust and Safety Engineering is responsible for the software and system architecture that defends the YouTube community of video creators, consumers and advertisers against a broad spectrum of abuse threats. This talk will provide some basic background on the organization, both in terms of composition and goals. We will then review two examples of extended abuse campaigns and how we managed them over time. My overall goal in this talk is to give researchers greater insight into the practical details of fighting abuse at YouTube scale.

Bio: J. Bradley Chen is a Principal Engineer at Google currently managing Systems Integrity at YouTube. Previously he managed YouTube's Revenue Data teams. Before that at Google he started the Native Client project and has also worked on cluster performance analysis projects. Prior to joining Google, he was Director of the Performance Tools Lab in Intel's Software Products Division. Chen served on the faculty of Harvard University from 1994-1998, conducting research in operating systems, computer architecture and distributed system, and teaching a variety of related graduate and undergraduate courses. He has published widely on the subjects of systems performance and computer architecture. Dr. Chen has bachelors and masters degrees from Stanford University and a Ph.D. from Carnegie Mellon University.

Full Presentation Session 1

• Disinformation Warfare: Understanding State-Sponsored Trolls on Twitter and Their Influence on the Web, Savvas Zannettou, Tristan Caulfield, Emiliano De Cristofaro, Michael Sirivianos, Gianluca Stringhini and Jeremy Blackburn. (Best Paper Award Winner)
• The Subtleties of Profanity in Abusive Language: Emergent Words and Meanings, Hongyu Gong, Alberto Valido, Katherine Ingram, Suma Bhat, Giulia Fanti and Dorothy Espelage
• Who Falls for Online Political Manipulation?, Adam Badawy, Kristina Lerman and Emilio Ferrara.
• An empirical study of malicious threads in security forums, Joobin Gharibshah, Zhabiz Gharibshah, Evangelos E. Papalexakis and Michalis Faloutsos

Short Presentation Session 1

• Understanding Cyberbullying on Instagram and Ask.fm via Social Role Detection, Hsien-Te Kao, Shen Yan, Di Huang, Nathan Bartley, Homa Hosseinmardi and Emilio Ferrara.
• Hi Doppelganger: Towards Detecting Manipulation in News Comments, Jan Pennekamp, Martin Henze, Oliver Hohlfeld and Andriy Panchenko.
• Detect Me If You Can: Spam Bot Detection Using Inductive Representation Learning, Seyed Ali Alhosseini, Bintareaf, Najafi and Meinel.

Full Presentation Session 2

• Robust Cyberbullying Detection with Causal Interpretation, Lu Cheng, Ruocheng Guo and Huan Liu
• Less is More: Semi-Supervised Causal Inference for Detecting Pathogenic Users in Social Media, Hamidreza Alvari, Elham Shaabani, Soumajyoti Sarkar, Ghazaleh Beigi and Paulo Shakarian.
• Characterizing Activity on the Deep and Dark Web, Nazgol Tavabi, Nathan Bartley, Andres Abeliuk, Sandeep Soni, Emilio Ferrara and Kristina Lerman
• "I was told to buy a software or lose my computer. I ignored it": A study of ransomware, Camelia Simoiu, Joseph Bonneau, Christopher Gates and Sharad Goel

Short Presentation Session 2

• Domain Generation Algorithms detection through deep neural network and ensemble, Shuaiji Li, Tao Huang, Zhiwei Qin, Fanfang Zhang and Yinhong Chang.
• Bot Electioneering Volume: Visualizing Social Bot Activity During Elections, Kai-Cheng Yang, Pik-Mai Hui and Filippo Menczer.
• Fraudulent Cryptocurrency Manipulations on Social Media, Mehrnoosh Mirtaheri, Sami Abu-El-Haija, Fred Morstatter, Greg Ver Steeg and Aram Galstyan.