This software is a tool to identify hardware Trojans present in integrated circuits (IC’s) by analyzing scanning electron microscope (SEM) images. A hardware Trojan is an intentional modification of an IC to leak valuable data or to disable the chip or its components. Integrated circuits are used in thousands of devices used in the internet of things (IOT,) civilian security, health care, automotive, military applications and more. Estimates predict the IC market to grow by 27 percent by 2023, reaching an estimated worth of $1.6 billion. Industry design houses commonly outsource the production of IC’s to foundries. In some cases, foundries will incorporate hardware Trojans into circuits to diminish their functionality or leak sensitive information later. Trojans hidden in the unused spaces of an IC during fabrication become a serious security threat to any device that uses ICs. Available hardware Trojan detection tools include electrical tests, such as run- and test-time monitoring, but are inefficient and often inaccurate. Run-time tests require extra CPU usage, power, and memory in order to function. Some test-time monitoring is sensitive to noise within the circuits and cannot detect small Trojans, while others require human-generated test vectors that have difficulty detecting large Trojans. Reverse engineering, a destructive test that requires the disassembly and detailed analysis of an IC, provides more accurate results, but is expensive and slow.
Researchers at the University of Florida have developed an image processing tool that identifies hardware Trojans present in integrated circuits in a shorter amount of time than available tools and with more accurate results. This tool uses an image processing algorithm to compare a SEM image of an IC that does not have any Trojans with another SEM image of an IC under evaluation. The evaluation completes in less than a day through a simple, accurate, image-based test.
Image processing tool to detect hardware Trojans in integrated circuits
This tool detects hardware Trojans by comparing SEM images of thinned IC’s. In the beginning of the evaluation process, a mechanical polisher thins the backside of the IC so electron beams from the SEM can penetrate it. A perfect IC that does not contain any hardware Trojans, known as a golden IC, also undergoes backside thinning. Next, the SEM captures images of both the golden IC at a high resolution and the IC under evaluation at a low resolution. Only one SEM image of the golden IC is necessary as a reference for evaluating of any number of IC’s. Gaussian and Median filters refine the SEM images to reduce noise and enhance their quality. Then, the Structural SIMilarity (SSIM) image processing index compares the images. SSIM identifies inconsistencies between the images of the two IC’s based on luminosity, contrast, and structure. Finally, a Trojan heat map made from the SSIM data and a machine learning based algorithm evaluates it to determine the presence and locations of hardware Trojans in the IC.