Search for: All records

Federated learning (FL) has emerged as a promising paradigm for secure distributed machine learning model training across multiple clients or devices, enabling model training without having to share data across the clients. However, recent studies revealed that FL could be vulnerable to data leakage and reconstruction attacks even if the data itself are never shared with another client. Thus, to resolve such vulnerability and improve the privacy of all clients, a class of techniques, called privacy-preserving FL, incorporates encryption techniques, such as homomorphic encryption (HE), to encrypt and fully protect model information from being exposed to other parties. A downside to this approach is that encryption schemes like HE are very compute-intensive, often causing inefficient and excessive use of client CPU resources that can be used for other uses. To alleviate this issue, this study introduces a novel approach by leveraging smart network interface cards (SmartNICs) to offload compute-intensive HE operations of privacy-preserving FL. By employing SmartNICs as hardware accelerators, we enable efficient computation of HE while saving CPU cycles and other server resources for more critical tasks. In addition, by offloading encryption from the host to another device, the details of encryption remain secure even if the host is compromised, ultimately improving the security of the entire FL system. Given such benefits, this paper presents an FL system named FedNIC that implements the above approach, with an in-depth description of the architecture, implementation, and performance evaluations. Our experimental results demonstrate a more secure FL system with no loss in model accuracy and up to 25% in reduced host CPU cycle, but with a roughly 46% increase in total training time, showing the feasibility and tradeoffs of utilizing SmartNICs as an encryption offload device in federated learning scenarios. Finally, we illustrate promising future study and potential optimizations for a more secure and privacy-preserving federated learning system. 

Pathways to the professoriate for women in computer science are narrow and fraught with barriers. These obstacles are further exacerbated at the intersections of race and gender. Black women (who make up 6.4% of the U.S. population) comprise only 1.1% of computer science undergraduate degrees and < 1% of computer science PhDs. Despite these paltry numbers, one computer science PhD program may have found the combination of factors necessary to widen the pathway by engaging in strategic recruitment, developing communities of practice, and providing strong mentorship for women of color in computer science. Guided primarily by intersectionality theory, social identity theory, and landscapes of practice, this single case study explored the experiences of Black women in pursuit of their doctorate in computer science at a predominantly white institution to answer the research questions: (1) How do Black women graduate students in computer science describe their computer science identity? (2) How do landscapes of practice influence computer science identity formation or salience of Black women in a computer science graduate program? Thematic analysis of this case revealed three common themes within their experiences: moments of impact, boundary spanning, and community residence. These themes, all of which revolve around ideas of community and support, are critical to understanding a key discovery of this study: why a sense of belonging, rather than identity salience (as much research suggests), was the best indicator of the women’s persistence. 

This study was designed to compare salary implications and employability of students who graduated with a Bachelor of Arts in Computer Science (BACS) – primarily distinguished by the removal of calculus and physics requirements from the traditional computer science curriculum versus those that graduated with a Bachelor of Science in Computer Science (BSCS). Given the numerous studies that identify gateway courses like calculus and physics as impediments to students’ persistence in engineering and computer science AND their impact on women and people of color, the removal of this barrier has incredible potential for broadening participation in computing. One university’s first cohort of BACS graduates (spring 2020) furnished a unique opportunity to compare student’s self-reported employment and salary information to their BSCS peers. The study consisted of institutional data and a survey targeting spring 2020, summer 2020, fall 2020 graduates from computer science, with data from n=134 recent graduates (BA n= 45, BS n=89). Preliminary results indicate there are no statistical significance in enrollment on the basis of gender nor job attainment; however, there is a statistical significance in enrollment on the basis of race/ethnicity and pay. The results of this work could either serve as a cautionary tale for institutions considering similar programs OR it could serve as the basis for a deeper, more critical review of the requirements currently in place in BSCS programs, nationally. Are calculus and physics courses required for prosperity in computing or are they simply a barrier to equity?