Ben Y. Zhao

Ben Y. Zhao

Auto-generated profile, based on publicly available information

Ben Y. Zhao

University of Chicago, Chicago, IL, USA

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 December 2024
Verified
Conference Paper
Open access
AHAnna Yoo Jeong Ha
Anna Yoo Jeong Ha
JPJosephine Passananti
Josephine Passananti
RBRonik Bhaskar
Ronik Bhaskar
SSShawn Shan
Shawn Shan
RSReid Southen
Reid Southen
BZBen Y. Zhao
Ben Y. Zhao
The advent of generative AI images has completely disrupted the art world. Distinguishing AI generated images from human art is a challenging problem whose impact is growing over time. A failure to address this problem allows bad actors to defraud individuals paying a premium for human art and companies whose stated policies forbid AI imagery. It is also critical for content owners to establish copyright, and for model trainers interested in curating training data in order to avoid potential model collapse. There are several different approaches to distinguishing human art from AI images, including classifiers trained by supervised learning, research tools targeting diffusion models, and identification by professional artists using their knowledge of artistic techniques. In this paper, we seek to understand how well these approaches can perform against today's modern generative models in both benign and adversarial settings. We curate real human art across 7 styles, generate matching images from 5 generative models, and apply 8 detectors (5 automated detectors and 3 different human groups including 180 crowdworkers, 3800+ professional artists, and 13 expert artists experienced at detecting AI). Both Hive and expert artists do very well, but make mistakes in different ways (Hive is weaker against adversarial perturbations while Expert artists produce higher false positives). We believe these weaknesses will persist, and argue that a combination of human and automated detectors provides the best combination of accuracy and robustness.


2 December 2024
Verified
Conference Paper
Open access
WDWenxin Ding
Wenxin Ding
CLCathy Y. Li
Cathy Y. Li
SSShawn Shan
Shawn Shan
BZBen Y. Zhao
Ben Y. Zhao
HZHaitao Zheng
Haitao Zheng
Recent works show that text-to-image generative models are surprisingly vulnerable to a variety of poisoning attacks. Empirical results find that these models can be corrupted by altering associations between individual text prompts and associated visual features. Furthermore, a number of concurrent poisoning attacks can induce "model implosion," where the model becomes unable to produce meaningful images for unpoisoned prompts. These intriguing findings highlight the absence of an intuitive framework to understand poisoning attacks on these models. In this work, we establish the first analytical framework on robustness of image generative models to poisoning attacks, by modeling and analyzing the behavior of the cross-attention mechanism in latent diffusion models. We model cross-attention training as an abstract problem of "supervised graph alignment" and formally quantify the impact of training data by the hardness of alignment, measured by an Alignment Difficulty (AD) metric. The higher the AD, the harder the alignment. We prove that AD increases with the number of individual prompts (or concepts) poisoned. As AD grows, the alignment task becomes increasingly difficult, yielding highly distorted outcomes that frequently map meaningful text prompts to undefined or meaningless visual representations. As a result, the generative model implodes and outputs random, incoherent images at large. We validate our analytical framework through extensive experiments, and we confirm and explain the unexpected (and unexplained) effect of model implosion while producing new, unforeseen insights. Our work provides a useful tool for studying poisoning attacks against diffusion models and their defenses.

10 February 2020 in Acta Geophysica
Verified
Journal Article
Closed access
PDF file
WZWeijia Zhang
Weijia Zhang
BZBen Zhao
Ben Zhao
XLXinyu Lou
Xinyu Lou

8 May 2021
Verified
Conference Paper
Open access
YCYuxin Chen
Yuxin Chen
ZYZhuolin Yang
Zhuolin Yang
RARuben Abbou
Ruben Abbou
PLPedro Lopes
Pedro Lopes
BZBen Y. Zhao
Ben Y. Zhao
HZHaitao Zheng
Haitao Zheng
We propose a novel modality for active biometric authentication: electrical muscle stimulation (EMS). To explore this, we engineered an interactive system, which we call ElectricAuth, that stimulates the user’s forearm muscles with a sequence of electrical impulses (i.e., EMS challenge) and measures the user’s involuntary finger movements (i.e., response to the challenge). ElectricAuth leverages EMS’s intersubject variability, where the same electrical stimulation results in different movements in different users because everybody’s physiology is unique (e.g., differences in bone and muscular structure, skin resistance and composition, etc.). As such, ElectricAuth allows users to login without memorizing passwords or PINs. ElectricAuth’s challenge-response structure makes it secure against data breaches and replay attacks, a major vulnerability facing today’s biometrics such as facial recognition and fingerprints. Furthermore, ElectricAuth never reuses the same challenge twice in authentications – in just one second of stimulation it encodes one of 68M possible challenges.