-
Integrating Accessibility in a Mobile App Development Course
Authors:
Jaskaran Singh Bhatia,
Parthasarathy P D,
Snigdha Tiwari,
Dhruv Nagpal,
Swaroop Joshi
Abstract:
The growing interest in accessible software reflects in computing educators' and education researchers' efforts to include accessibility in core computing education. We integrated accessibility in a junior/senior-level Android app development course at a large private university in India. The course introduced three accessibility-related topics using various interventions: Accessibility Awareness…
▽ More
The growing interest in accessible software reflects in computing educators' and education researchers' efforts to include accessibility in core computing education. We integrated accessibility in a junior/senior-level Android app development course at a large private university in India. The course introduced three accessibility-related topics using various interventions: Accessibility Awareness (a guest lecture by a legal expert), Technical Knowledge (lectures on Android accessibility guidelines and testing practices and graded components for implementing accessibility in programming assignments), and Empathy (an activity that required students to blindfold themselves and interact with their phones using a screen-reader). We evaluated their impact on student learning using three instruments: (A) A pre/post-course questionnaire, (B) Reflective questions on each of the four programming assignments, and (C) Midterm and Final exam questions. Our findings demonstrate that: (A) significantly more ($p<.05$) students considered disabilities when designing an app after taking this course, (B) many students developed empathy towards the challenges persons with disabilities face while using inaccessible apps, and (C) all students could correctly identify at least one accessibility issue in the user interface of a real-world app given its screenshot, and 90% of them could provide a correct solution to fix it.
△ Less
Submitted 12 October, 2022;
originally announced October 2022.
-
Exploiting and Defending Against the Approximate Linearity of Apple's NeuralHash
Authors:
Jagdeep Singh Bhatia,
Kevin Meng
Abstract:
Perceptual hashes map images with identical semantic content to the same $n$-bit hash value, while mapping semantically-different images to different hashes. These algorithms carry important applications in cybersecurity such as copyright infringement detection, content fingerprinting, and surveillance. Apple's NeuralHash is one such system that aims to detect the presence of illegal content on us…
▽ More
Perceptual hashes map images with identical semantic content to the same $n$-bit hash value, while mapping semantically-different images to different hashes. These algorithms carry important applications in cybersecurity such as copyright infringement detection, content fingerprinting, and surveillance. Apple's NeuralHash is one such system that aims to detect the presence of illegal content on users' devices without compromising consumer privacy. We make the surprising discovery that NeuralHash is approximately linear, which inspires the development of novel black-box attacks that can (i) evade detection of "illegal" images, (ii) generate near-collisions, and (iii) leak information about hashed images, all without access to model parameters. These vulnerabilities pose serious threats to NeuralHash's security goals; to address them, we propose a simple fix using classical cryptographic standards.
△ Less
Submitted 28 July, 2022;
originally announced July 2022.
-
Evolution Gym: A Large-Scale Benchmark for Evolving Soft Robots
Authors:
Jagdeep Singh Bhatia,
Holly Jackson,
Yunsheng Tian,
Jie Xu,
Wojciech Matusik
Abstract:
Both the design and control of a robot play equally important roles in its task performance. However, while optimal control is well studied in the machine learning and robotics community, less attention is placed on finding the optimal robot design. This is mainly because co-optimizing design and control in robotics is characterized as a challenging problem, and more importantly, a comprehensive e…
▽ More
Both the design and control of a robot play equally important roles in its task performance. However, while optimal control is well studied in the machine learning and robotics community, less attention is placed on finding the optimal robot design. This is mainly because co-optimizing design and control in robotics is characterized as a challenging problem, and more importantly, a comprehensive evaluation benchmark for co-optimization does not exist. In this paper, we propose Evolution Gym, the first large-scale benchmark for co-optimizing the design and control of soft robots. In our benchmark, each robot is composed of different types of voxels (e.g., soft, rigid, actuators), resulting in a modular and expressive robot design space. Our benchmark environments span a wide range of tasks, including locomotion on various types of terrains and manipulation. Furthermore, we develop several robot co-evolution algorithms by combining state-of-the-art design optimization methods and deep reinforcement learning techniques. Evaluating the algorithms on our benchmark platform, we observe robots exhibiting increasingly complex behaviors as evolution progresses, with the best evolved designs solving many of our proposed tasks. Additionally, even though robot designs are evolved autonomously from scratch without prior knowledge, they often grow to resemble existing natural creatures while outperforming hand-designed robots. Nevertheless, all tested algorithms fail to find robots that succeed in our hardest environments. This suggests that more advanced algorithms are required to explore the high-dimensional design space and evolve increasingly intelligent robots -- an area of research in which we hope Evolution Gym will accelerate progress. Our website with code, environments, documentation, and tutorials is available at http://evogym.csail.mit.edu.
△ Less
Submitted 24 January, 2022;
originally announced January 2022.
