Browse Prior Art Database

Intrinsic ID Bitmap Authentication System Using Production Bitmap Images

IP.com Disclosure Number: IPCOM000236843D
Publication Date: 2014-May-19
Document File: 2 page(s) / 52K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a tool to accelerate the authentication process of large random strings derived from intrinsic properties of a chip based on recognition thresholds. The solution comprises the automatic processing and storage of all "golden" bitmap images from production, an ID table to cross-reference stored images, a fail counter, and an algorithm that compares the number of overlapping failing bit addresses between the field device and every "golden' bitmap in the ID table

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 2

Intrinsic ID Bitmap Authentication System Using Production Bitmap Images

It is known that random and unique chip identifiers (IDs) can be extracted from intrinsic chip fingerprints, in particular electrical signature fingerprints. Such Physically Unclonable Functions (PUFs) have been demonstrated using various types of chips exploiting mostly process variability. These intrinsic chip IDs can be a viable and secure way to protect against counterfeit of devices in the field, because the chip IDs are extremely difficult to copy and are unique to each piece of hardware.

Authentication of these intrinsic chip IDs, for use in a manufacturing line or by a customer, is challenging. The larger the number of bits available to use as an intrinsic chip ID is, the more unique the ID is. However, that creates a problem for authentication, as it requires more computations when authenticating the part.

Additionally, the intrinsic chip IDs are typically not stable enough and must be recognized by means of fuzzy pattern recognition. The computational power involved in authenticating long bit strings (12KB to 200MB of uncompressed data typical for the eDRAM approach) in a fuzzy manner may result in the need to check the existence of the bit string provided by the customer against the bit strings stored for every part in the manufacturer's database. For example, with 1 Million parts and typical 20Mbit maps (embedded Dynamic Random Access Memory (eDRAM) case) this can be time consuming. If this is part of a secure exchange of information such as a credit card transaction, then the customer could wait for hours for the transaction to end, which is unacceptable.

Therefore, methods for reducing intrinsic chip ID authentication time are necessary for practical use.

The unique bit strings representing the chip fingerprint IDs can be stored in binary form as bitmaps, with the immediate advantage of using significantly less storage resources than literal bit strings (1KB to 6MB for Tiff images of 20Mbit eDRAM IDs). The binary representation also presents the opportunity to do binary operations on the data, accelerating the recognition process by orders of magnitude. A non-optimal implementation of the bitmap recognition results in a full comparison of two 20Mbit bitmap images in 77ms on average on a personal computer (PC). Parallel...