IBM’s new z16 mainframe: real-time fraud detection at scale

Today, IBM introduces the z16 Dayden, the successor to the z15 mainframe that launched three years ago. The z16 is based on the new Telum processor, it launches on May 31, 2022 with the usual value for money improvements and brings two qualitative improvements: on-chip AI accelerators and quantum-proof computing.

According to IBM, the z in the new mainframe’s product name represents the zero-downtime design goal, the 16 for the 16th generation of the forward-compatible mainframe introduced in 1964/360. Without going into too much technical detail, Ross Mauri, general manager responsible for IBM’s systems business, presented the two most important innovations at a press conference: AI inference within transactions at scale and the first quantum security features of IBM. the industry to protect customer data.

The on-chip AI accelerator built into the Telum processor offers latency-optimized inference, which should enable real-time transaction analysis at scale, for example, for mission-critical workloads such as credit cards, healthcare, and financial transactions. Until now, it has not been possible to use real-time large-scale deep learning models due to latency issues, because real-time data, for example a credit card transaction, is outsourced and processed using a special data processing system. AI. This means, according to Mauri, that fraud detection models are now used in less than 10 percent of high-volume transactions; a significant portion of fraud goes undetected.

With the z16, for the first time, banks can scan large-scale transactions for fraud as they run — up to 300 billion requests per day with a latency of just one millisecond, Mauri said. These numbers were extrapolated from initial tests running local inference operations on a z16 LPAR with 48 IFLs and 128 GB of memory on Ubuntu 20.04 (SMT mode) using a synthetic credit card fraud detection model and the built-in Accelerator for AI. The benchmark used 8 parallel threads for this, each of which was connected to the first core of a different chip.

According to Elpida Tzortzatos, CTO, IBM Systems, AI, AI will reduce the time it takes to detect a fraudulent credit card transaction or speed up a loan approval from minutes to microseconds. For consumers, this could mean spending less time and energy trying to avoid fraudulent credit card transactions. For both merchants and card issuers, this could mean reduced lost revenue as customers avoid the frustration of false declines and switch cards for future transactions. Mauri names other fields of application of this AI technology:

  • Credit Approval – Speed ​​up consumer or business credit approval
  • Clearing and Settlement – ​​Identify which trades and/or transactions may be high risk prior to settlement
  • Federated training for retailers: to better model the risk of fraud and theft.

Building on IBM technologies such as Pervasive Encryption and Confidential Computing, IBM z16 goes one step further by protecting data from future threats that may evolve with advances in quantum computing. A z16 uses so-called grid-based cryptography, for example for secure boot. This means that attackers cannot inject malware into the boot process to control the system during startup.

In addition, the Crypto Express 8S (CEX8S) coprocessor provides customers with classical and quantum proof cryptographic technologies in hardware to enforce confidentiality, integrity and non-repudiation of information. The map provides secure quantum APIs to access secure quantum algorithms. The secure boot process and secure quantum cryptography should already help prevent future threats related to quantum computers, including attacks like Harvest Now or Decrypt Later, which can lead to extortion, loss of intellectual property, or disclosure of sensitive data.

[Update 5. April 2022, 08:05 Uhr:] IBM has now provided an overview of technical data and information about Linux on the z16,


More from iX Magazine

More from iX Magazine


More from iX Magazine

More from iX Magazine


(fo)

to the home page

.

Leave your vote

Leave a Comment

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.