Implementing RAVEN AI Surveillance at a Major Shopping Mall: Reducing Threats and Enhancing Safety

A major shopping mall in Nairobi, one of the city's largest retail destinations, faced increasing security challenges from high foot traffic and potential terrorism risks. Traditional CCTV systems were inadequate for proactive threat detection in such a dynamic environment. Partnering with Boldstreet Partners, the mall implemented RAVEN—a sophisticated AI surveillance system that integrates computer vision, audio detection, thermal sensing, and blockchain for tamper-proof evidence. This case study details the challenges faced, the tailored solution, implementation process, measurable results, and key learnings, providing a blueprint for other Kenyan malls and public venues seeking to elevate their security posture. We elaborate on various use cases where RAVEN proved instrumental in real-world scenarios, such as detecting unattended objects, identifying suspicious behaviors, and managing crowd anomalies.

Challenge

• High-volume footfall in the mall's corridors and open spaces overwhelmed manual monitoring, leading to missed suspicious activities like unattended bags or unusual crowd behaviors.
• Slow response times to incidents due to reliance on human triage, which could take minutes in critical situations where seconds matter.
• Evidence integrity issues in post-incident investigations, with footage vulnerable to tampering and disputes over chain of custody delaying insurance claims and legal proceedings.

Solution

• RAVEN's multi-modal detection fused visual, audio, and thermal data to generate high-confidence alerts, reducing false positives and enabling proactive interventions. For example, in high-traffic areas, audio cues like raised voices combined with visual anomalies trigger immediate alerts.
• Edge-first processing ensured low-latency detection while keeping raw footage local, with only metadata sent to the central dashboard for privacy compliance.
• Automated playbooks for high-priority alerts, integrated with the mall's security protocols, to dispatch teams and notify authorities instantly. This includes scenarios like perimeter breaches or abnormal heat signatures indicating potential hazards.
• Blockchain-anchored evidence logs created immutable records for every incident, streamlining forensics and legal processes, ensuring that evidence from events like theft or altercations is court-admissible.

Implementation

• Conducted a site audit of existing 150+ cameras, identifying gaps and integrating audio/thermal sensors in high-risk zones like entrances and food courts.
• Piloted in 4 zones over 2 months, refining AI models with local data to account for Nairobi's unique crowd patterns and environmental factors.
• Trained security staff on the new dashboard and playbooks, with full rollout across the mall in under 4 months.
• Integrated with existing control room systems and emergency services for seamless operations.

Results

• 40% reduction in response times to potential threats, with alerts now triggering actions in under 30 seconds.
• 75% drop in false alarms, freeing up security personnel for genuine issues and reducing operational fatigue.
• Successful prevention of 3 potential incidents, including unattended objects flagged and cleared without evacuation.
• Faster claims processing, with blockchain evidence accepted in court, cutting resolution time by 60%.

Key Learnings

• Customizing AI models to local contexts, like Kenyan crowd behaviors, significantly improves accuracy.
• Privacy-by-design features, such as face blurring and signage, build public trust and ensure regulatory compliance.
• Ongoing training and playbook iterations are essential for maximizing system effectiveness.