Community Partners in Residence (CPR) Program – Meeting 5

On Tuesday, February 6, 2024, the NSF Center for Smart StreetScapes (CS3) held its fifth Community Partners in Residence (CPR) Program workshop. The session was convened and led once again by Ester Fuchs, Chief Social Impact Officer of the CS3 and Professor of International and Public Affairs and Political Science at Columbia University. The focus of this week’s workshop was on the subjects of security, privacy, and fairness as they pertain to emerging streetscape technology.

Roxana Geambasu, an assistant professor of computer science at Columbia University, opened the session by addressing privacy, fairness, and transparency issues in streetscape technologies. She acknowledged the significant potential of these technologies but also highlighted risks such as compromised sensors, potential hacker attacks, and the misuse of city data. Geambasu proposed a comprehensive approach to enhance security, privacy, and fairness through community engagement, emphasizing the need for a programming framework for future streetscape applications.

The presentation delved into the tradeoff between accuracy and privacy in real-time data feeding to streetscape applications, introducing the concept of “anonymization” as the first level of data protection. This process involves edge computers for object detection and tracking, providing real-time anonymized data to applications. Geambasu acknowledged the drawbacks of anonymization, leading to the discussion of an advanced protection level that involves sharing noisy aggregates over the dataset instead of raw data. This method prevents accurate reconstruction of individual entries, striking a balance between privacy and meaningful tradeoffs for streetscape applications. Geambasu concluded by emphasizing the crucial role of effective researcher-community engagement in navigating this tradeoff, providing specific and actionable guidance.

Following Prof. Geambasu’s presentation, the floor was opened to the CPRs for questions. Participants seemed primarily concerned with the security and ownership of data, with a majority of questions being devoted to ascertaining who would control the data generated by smart streetscape infrastructure, as well as how demands to access the data by third parties would be handled. In addition, CPRs floated questions regarding the anonymization of data and protection of individual privacy. Prof. Geambasu’s responses generally served to assure CPRs that data security and anonymization would occur as soon as possible within the pipeline, minimizing risk for people being observed. CPRs were also informed of the existence of Columbia University’s independent internal review board for the project.

The second presentation by Salvatore Stolfo, a computer science professor at Columbia University, highlighted current security challenges, including the risks associated with compromised IoT sensors and edge server hacking. He emphasized the need for policies and transparency in handling data collected by third parties and underscored the importance of further research on AI failures. The CPR survey results revealed concerns about data misuse, particularly with companies, and discomfort with technologies like facial recognition and AI. Stolfo also discussed the community’s interest in understanding how the technology is engaged with community application and how they can be abused. Stolfo emphasized the significance of finding the right tradeoff, acknowledging the benefits of technology in preventing crimes and terrorist attacks. He presented ongoing research addressing the protection of physically deployed sensors and the overall system, including centralized command and control for managing security technology and self-protecting streetscape security systems. Additionally, he highlighted research efforts focused on AI failures and trustworthiness to enhance identity authentication, scene analysis, and situation awareness accuracy.

After Prof. Stolfo’s presentation, CPRs raised questions regarding the accuracy of facial recognition technology when used on people with darker skin and in a crowd setting, as well as the identification of legitimate data. Their concerns were acknowledged, with Prof. Stolfo reassuring the audience that facial recognition technology is rather robust in certain contexts. He admitted that there are ongoing difficulties in applying the technology to finding a subset of individuals within a set of many faces, and also warned participants of the difficulties in discerning legitimate software from malicious imitations. He also acknowledged the possibility that a well-intentioned actor may accidentally cause harm to a system.

The meeting transitioned into breakout sessions, where the groups of CPRs shared today’s insights facilitated by CS3 staff. A prevalent concern raised by the majority was the privacy risk associated with valuable data, even if being anonymized. They expressed the need for a deeper understanding and discussion regarding the tradeoff between privacy and functionality of data. Facilitators initiated discussion on challenges of the streetscapes impacted by privacy security and transparency. In active discussions, CPRs highlighted the demand for protective measures for physical technology, transparency in politics and procedures, and the trustworthiness of the systems. They emphasized the importance of decentralized control of data acquisition, addressing environmental and contextual biases in facial recognition, and mitigating the risk of hackers maliciously controlling streetscapes. Additionally, there was a collective awareness of the significance of education required to find a balance between functionality and privacy, with CPRs proposing discussion among community reporters on TV and at community board meetings to further address this issue.

CPR workshops are invite-only. If you are interested in attending a workshop or learning more about CS3’s community engagement process, please contact our team at streetscapes@columbia.edu.