MlondaCam: Designing Context-Aware Smart Cameras for Supporting Domestic Security and Privacy in Malawian Homes

Paper Title

MlondaCam: Designing Context-Aware Smart Cameras for Supporting Domestic Security and Privacy in Malawian Homes

Publication Info

• Topic area: Context-aware smart cameras for domestic security and privacy in sub-Saharan Africa.
• Keywords: Context-aware cameras, domestic security, privacy, patriarchal norms, Malawi, HCI4D, feminist HCI, surveillance, gender equity, smart home technologies.

Background and Problem

• Problem / challenge: Existing smart cameras, while effective for security, risk exacerbating patriarchal norms and violating privacy within households, especially in regions like Malawi where gendered power dynamics are prevalent. Research on domestic security technologies has largely focused on Western contexts, leaving gaps in understanding their use in sub-Saharan Africa.
• Significance: Addressing domestic insecurity is critical in Malawi, where over 70% of urban crimes involve theft, and patriarchal norms often limit women’s agency. Designing privacy-preserving technologies can mitigate these issues while enhancing security.
• Motivation and related work: Prior studies have examined smart cameras in Western contexts and explored privacy-preserving designs, but little attention has been given to their adoption in patriarchal and resource-constrained environments. Previous research in Malawi suggested context-aware cameras that disable indoors could protect women’s privacy, but these designs had not been deployed or studied in real households.

Solution

• Proposed approach: MlondaCam, a context-aware smart camera that automatically disables itself indoors while remaining active outdoors, addressing both security and privacy concerns in Malawian households.
• Novelty:
  1. Deployment and evaluation of a context-aware camera in a resource-constrained, patriarchal context.
  2. Empirical insights into how households in Malawi appropriate security cameras for everyday life and navigate privacy-security tensions.
  3. Design considerations for context-aware technologies that balance security needs with privacy and gender equity.
• Procedure and key techniques:
  1. Developed MlondaCam using a Raspberry Pi 4, Arducam module, and PIR sensor, with a MobileNet-based classifier to distinguish indoor and outdoor environments.
  2. The system records video when motion is detected outdoors and disables recording indoors, with notifications sent via a mobile app.
  3. Deployed the system in 15 households in Lilongwe, Malawi, for four weeks, collecting data through semi-structured interviews, diaries, and observations.

Results

• Concrete findings:
  • MlondaCam achieved 97% training accuracy and 98% validation accuracy in distinguishing indoor and outdoor environments, with an F1 score of 0.96 on unseen test data.
  • Participants used the camera for diverse purposes, including deterring theft, protecting crops, managing shared resources, and uncovering unauthorized activities.
  • Women valued the indoor disable feature as protection against spousal surveillance, while men often saw it as a limitation.
• Advantage over baselines:
  • MlondaCam’s context-aware functionality directly addressed patriarchal misuse risks, unlike conventional smart cameras that lack such safeguards.
  • The system supported practical, everyday uses beyond burglary prevention, aligning with participants’ routines and needs.
• Experiments / evaluation:
  • Deployment in 15 households with diverse compositions and occupations.
  • Data collection through semi-structured interviews, diaries, and observations over a four-week period.
  • Analysis revealed tensions around visibility, authority, and privacy, shaped by gendered hierarchies and infrastructural constraints.
• Limitations and future work:
  • False alerts and limited camera coverage frustrated participants, highlighting the need for improved filtering and multi-camera setups.
  • Deployment challenges included patriarchal resistance, smartphone compatibility issues, theft of equipment, and infrastructural fragility (e.g., unreliable internet and power).
  • Future work should explore modular, portable designs, finer-grained privacy controls, and co-design approaches with marginalized household members.

Summary

This study introduced and evaluated MlondaCam, a context-aware smart camera designed to balance domestic security and privacy in Malawian households. The system’s automatic indoor disable feature was celebrated by women as protection against patriarchal surveillance but seen as a limitation by men. Participants appropriated the camera for diverse everyday uses, including protecting crops and managing shared resources, revealing how security needs intersect with livelihood and dignity. While the deployment highlighted the potential of bounded surveillance, it also surfaced tensions around visibility, authority, and infrastructural fragility. The findings underscore the importance of designing flexible, context-sensitive technologies that account for gendered power dynamics and resource constraints.