A network-based approach to counterfeit detection

Counterfeit devices are wreaking havoc in the industry today and are causing billions of dollars of loss in revenue to companies. These devices usually have fake components or have components that are re-marked as better ones. Illegitimate workshops even manufacture counterfeit devices on a large scale. Hence, counterfeit detection is of utmost importance. However, most counterfeit detection techniques today have complex and expensive setups that are slow in generating results. Some methods are also destructive to the device under test, and can hence only be applied to a small portion of the suspect devices. Most methods are also usually intended to target a specific set of devices. In this paper, we propose a simple and cheap technique of counterfeit detection, which we believe is a first-of-its-kind, network-based solution. Being network-based, it can be used to swiftly test a broad range of networked devices. The technique only monitors the network traffic of the device, therefore it is non-destructive. We first illustrate the general efficacy of the technique using FPGAs. Next, we show that lower-end processors (i.e., Core i3) in real systems can be differentiated from higher-end processors (i.e., Core i7) based on the host node's network traffic. This technique is effective against devices with counterfeit components or even with legitimate, but re-marked (as higher capacity) components. Using a neural network based classifier, we show classifier recall values (i.e., the ratio of the number of true positives to the sum of the number of true positives and false negatives) of up to 78.7% using traffic captures of 2,500 packets.