According to Palo Alto chief product officer Lee Klarich, the security industry has been able to reduce the time it takes to deliver detection signatures from around 24 hours to five minutes or so. But attackers have responded by creating systems that can rapidly and automatically generate large numbers of variants. In this context, 'large numbers' means hundreds of thousands per day.
Palo Alto's use of inline machine learning blocks up to 95% of these malware and phishing variants, he said. But the remaining 9% means "signatures still play an important role".
The company can now deliver a protection signature to customers within around five seconds of the first detection of a new threat, Klarich said, which is "just about real time."
Machine learning is also brought to bear on the problem of securing IoT devices.
"IoT is infiltrating networks all over the place," he said, and "securing them can be hard."
Where other approaches require new infrastructure, Palo Alto applies ML within NGFW to provide visibility of IoT devices on the network, identifying the device type and the software version it is running. This combines technology from Zingbox, which Palo Alto acquired in 2019, with its homegrown App-ID technology.
It then reports any known vulnerabilities, allowing the organisation to prioritise the updates or other remedial action.
ML is also used to identify anomalous IoT behaviour (such as changes in inbound or outbound data flows, and attempts to connect to applications or servers for the first time), and to recommend security policies for IoT devices.
"This is the start of a bigger future," Klarich said, implying that ML-defined security policies will be extended to other types of devices.
While these ML-based capabilities headlined Palo Alto's announcement, more than 70 new capabilities are delivered with PAN-OS 10.0. "This is a big release," he said.
They include a high-performance processing card for the hardware NGFW, a containerised edition of NGFW (CN-Series) designed specifically for Kubernetes environments, improved SSL decryption (including TLS 1.3 support), high-availability clusters of up to 16 NGFWs, and the ability to quarantine devices attempting to join the network.
"Thirteen years ago, we completely changed network security when we created the Next-Generation Firewall," said Palo Alto Networks founder and chief technology officer Nir Zuk.
"As enterprise networks are widening – with hybrid clouds, IoT devices and home offices – and attacks rapidly and automatically evolve, we again need a radical new approach to cybersecurity. PAN-OS version 10.0 ushers in the world's first ML-Powered NGFW, which is continuously learning and proactively improving security across multiple fronts, so security professionals don't just keep up but get ahead."
PAN-OS 10.0 is slated for release in mid-July, and will be available to all current customers with support contracts. The IoT features will be delivered as a separate subscription.