Ping Identity Agentic Access Model

How do bots authoritatively access your resources? The Ping Identity agentic access model documents one way to control access.

The ultimate guide?

Ping Identity wrote a page called “The Ultimate Guide to Identity for AI.” Ultimate? I guess a marketer wrote it.

Ping Identity notes that today’s agents are “navigating systems, initiating transactions, and making decisions—autonomously or under delegated authority.” Note that word “autonomously.” The system must control agentic access so we don’t get “agents gone wild” stories.

"Agents gone wild" without the Ping Identity agentic access model. Google Gemini.
“Agents gone wild” without the Ping Identity agentic access model. Google Gemini.

Controlling agentic access

To prevent such robotic debauchery, Ping Identity uses several mechanisms, including OAuth 2.0, DCR, mTLS, and assertion grants to control those pesky little bots. In addition, Ping Identity promotes four pillars of governance.

First: Delegate, don’t impersonate

You don’t want the agents to pretend they are humans and do whatever. Instead, the system issues “scoped, temporary tokens that clearly define what an agent is authorized to do, and on whose behalf.” So the system knows the human owner, and the authorizations that prohibit that expensive champagne purchase.

Second: Enforce least privilege

Just as least privilege is critically important when letting humans do and not do things, it’s also critically important for the bots. After all, authorizations apply to human and non-human identities alike (attribute based access control), and the system restricts them as needed.

Third: Maintain human oversight

We know this is important, but how do we do it? By designing systems that include humans in the loop. I’ve discussed that to death.

Fourth: Make every action auditable

Finally, make sure you know what everyone and everything did.

The risks are real

I don’t need to tell you that the risks of uncontrolled agents are real. Maybe they won’t launch World War III, but they can do plenty of damage. As Ping Identity reminds us, “Without clear identity and access boundaries, AI agents can become untraceable actors in systems that were never designed to govern or monitor them.”

That’s why Ping Identity and other vendors are implementing ways to control the new world in which humans and agents (hopefully) work together.

And old-timers such as me are adjusting to the new reality. If we don’t, we’re in trouble.

Google Gemini.

Deepfakes Slipping Through the Silos

(Imagen 4)

Sometimes common sense isn’t enough to stop deepfake fraud. Marc Ricker of iValt asserrts that a unified response helps also.

“Too often, network teams focus on availability, while security teams chase threats after the fact. That separation creates gaps — gaps that attackers exploit.”

Ricker’s solution:

“iVALT unifies remote access and identity security through:

Instant, passwordless biometric authentication

AI-resistant technology that stops deepfake and synthetic identity fraud”

iVALT trumpets the use of 5 factors: device ID, biometrics, geolocation, time window, and “app code.” 

  • I was curious which biometric modalities and vendors iVALT supported, so I looked it up. 
  • iVALT appears to use PingOne DaVinci, which orchestrates everything. 
  • The only biometrics specifically mentioned by iVALT are those captured on a mobile phone.
  • But it’s unclear to me whether these are the biometrics captured by the phone’s operating system (for example, TouchID or FaceID on iOS), third party biometrics, or all of the above.

Of course, most people don’t care about the minutiae of supported biometric modalities. 

But some do…because all biometric algorithms do NOT provide the same accuracy or performance.

Contraction

While the words “consolidation” and “contraction” have a similar sound and are often linked, they are actually two separate conditions, as you can see in the identity/biometric industry.

  • Consolidation occurs when separate entities become one. Ping Identity and ForgeRock (Ping Identity). Sagem Morpho and Motorola’s Biometric Business Unit (MorphoTrak). Digital Biometrics and Identix and Viisage and Visionics and Iridian and ComnetiX and don’t forget the ID part of Digimarc and many others (L-1 Identity Solutions).
  • Contraction occurs when an existing entity becomes smaller. Hikvision’s reported layoff of 1,000 employees is a recent and relevant example.

“Ah, but Hikvision is a special case,” you may be saying. “They’re linked to human rights abuses and sanctioned by Western governments. Many identity/biometric players are not sanctioned.”

But I’m not hearing loud celebrations from these other firms.

I’ve privately heard three separate stories, one of which I just heard on Monday, involving major identity/biometric companies. All three stories involve firms that are not sanctioned. In all three cases the firms perform major business with Western governments. And all three stories involve contraction which would have been unthinkable a mere five years ago.

Not too long ago I compiled a list of four significant events that positively impacted the identity/biometric industry. That list included 9/11, the Boston Marathon bombings, Apple’s Touch ID, and COVID.

I’m starting to wonder whether that last event was, in the long term, a net positive or a net negative.

(Tumbleweed image public domain)