Next-level network observability with AI-powered eBPF: a live discussion

“eBPF wasn’t designed for telecom networks, but by adapting it with AI, we’ve created a breakthrough in anomaly detection and security monitoring.” – David Soldani, SVP of Next-Generation Advanced Research, Rakuten Mobile

Anomaly detection and security in telco networks is getting harder as complexity scales. Teams typically put up with known blind spots, high volumes of false positives and the challenge of rigid architectures that can’t adapt to evolving threats. This is changing with help from a technology that wasn’t even designed with telecom in mind.

On this week’s episode of Zero-Touch Live, Rakuten Mobile SVP of Next-Generation Advanced Research David Soldani reveals how Rakuten Mobile is using eBPF (Extended Berkeley Packet Filter) and AI-driven anomaly detection to transform observability practices at scale, cutting false positives and detecting threats in real time.

The replay of our discussion is available below.

eBPF a big breakthrough with real results

Traditional telco network monitoring solutions tend to introduce inefficiency by duplicating packets, inserting probes and mirroring traffic.

"Operators have tried to improve detection for years but rules-based approaches simply don’t scale. We needed a new way to monitor security threats without overwhelming teams with false alarms," David explained.

Rakuten is using eBPF for real-time, kernel-level monitoring to:

• Detect anomalies at the transport layer without duplicating traffic.
• Prevent intrusions in real time and reduce security blind spots.
• Feed AI models with deeper network insights than traditional monitoring tools.

Importantly, Rakuten Mobile has accomplished these objectives without having to modify the network’s underlying infrastructure.

"Think of eBPF like being able to upgrade the engine of a car while it’s driving," said David.

Rakuten Mobile integrated AI-driven anomaly detection by applying models like DBSCAN clustering and variational autoencoders (VAEs) to reduce false positives, detect zero-day threats without relying on pre-defined rules and enabling self-learning security models.

While traditional monitoring approaches flood teams with alerts, AI-powered advances can extract real threats from the noise to make networks more secure and efficient.

An eBPF roadmap for telcos

eBPF adoption in telecom networks solves an increasing need as operators move toward cloud-native, containerized network architectures. David shared his advice for telcos eager to modernize:

• Adopt eBPF for real-time, low-overhead anomaly detection.
• Use AI-driven models to replace rigid, rules-based security approaches.
• Integrate automation to reduce manual security interventions.

Watch the replay of our discussion now to learn how eBPF and AI are reshaping telecom security.

‍