Today, face editing is widely used to refine/alter photos in both professional and recreational settings. Yet it is also used to modify (and repost) existing online photos for cyberbullying. Our work considers an important open question: 'How can we support the collaborative use of face editing on social platforms while protecting against unacceptable edits and reposts by others?' This is challenging because, as our user study shows, users vary widely in their definition of what edits are (un)acceptable. Any global filter policy deployed by social platforms is unlikely to address the needs of all users, but hinders social interactions enabled by photo editing. Instead, we argue that face edit protection policies should be implemented by social platforms based on individual user preferences. When posting an original photo online, a user can choose to specify the types of face edits (dis)allowed on the photo. Social platforms use these per-photo edit policies to moderate future photo uploads, i.e., edited photos containing modifications that violate the original photo's policy are either blocked or shelved for user approval. Realizing this personalized protection, however, faces two immediate challenges: (1) how to accurately recognize specific modifications, if any, contained in a photo; and (2) how to associate an edited photo with its original photo (and thus the edit policy). We show that these challenges can be addressed by combining highly efficient hashing based image search and scalable semantic image comparison, and build a prototype protector (Alethia) covering nine edit types. Evaluations using IRB-approved user studies and data-driven experiments (on 839K face photos) show that Alethia accurately recognizes edited photos that violate user policies and induces a feeling of protection to study participants. This demonstrates the initial feasibility of personalized face edit protection. We also discuss current limitations and future directions to push the concept forward.

Advances in deep learning have introduced a new wave of voice synthesis tools, capable of producing audio that sounds as if spoken by a target speaker. If successful, such tools in the wrong hands will enable a range of powerful attacks against both humans and software systems (aka machines). This paper documents efforts and findings from a comprehensive experimental study on the impact of deep-learning based speech synthesis attacks on both human listeners and machines such as speaker recognition and voice-signin systems. We find that both humans and machines can be reliably fooled by synthetic speech, and that existing defenses against synthesized speech fall short. These findings highlight the need to raise awareness and develop new protections against synthetic speech for both humans and machines.

Biases in language influence how we interact with each other and society at large. Language affirming gender stereotypes is often observed in various contexts today, from recommendation letters and Wikipedia entries to fiction novels and movie dialogue. Yet to date, there is little agreement on the methodology to quantify gender stereotypes in natural language (specifically the English language). Common methodology (including those adopted by companies tasked with detecting gender bias) rely on a lexicon approach largely based on the original BSRI study from 1974.

Malicious crowdsourcing forums are gaining traction as sources of spreading misinformation online, but are limited by the costs of hiring and managing human workers. In this paper, we identify a new class of attacks that leverage deep learning language models (Recurrent Neural Networks or RNNs) to automate the generation of fake online reviews for products and services. Not only are these attacks cheap and therefore more scalable, but they can control rate of content output to eliminate the signature burstiness that makes crowdsourced campaigns easy to detect.

For millions of workers, online job listings provide the first point of contact to potential employers. As a result, job listings and their word choices can significantly affect the makeup of the responding applicant pool. Here, we study the effects of potentially gender-biased terminology in job listings, and their impact on job applicants, using a large historical corpus of 17 million listings on LinkedIn spanning 10 years. We develop algorithms to detect and quantify gender bias, validate them using external tools, and use them to quantify job listing bias over time. We then perform a user survey over two user populations (N 1=469 , N 2=273 ) to validate our findings and to quantify the end-to-end impact of such bias on applicant decisions. Our findings show gender-bias has decreased significantly over the last 10 years. More surprisingly, we find that impact of gender bias in listings is dwarfed by our respondents' inherent bias towards specific job types.