Trojan

About: Trojan is a research topic. Over the lifetime, 2028 publications have been published within this topic receiving 33209 citations.

Detecting Hardware Trojans Inserted by Untrusted Foundry Using Physical Inspection and Advanced Image Processing

Abstract: Hardware Trojans are malicious changes to the design of integrated circuits (ICs) at different stages of the design and fabrication process. Different approaches have been developed to detect Trojans namely non-destructive and destructive testing. However, none of the previously developed methods can be used to detect all types of Trojans as they suffer from a number of disadvantages such as low speed of detection, low accuracy, low confidence level, and poor coverage of Trojan types. Majority of the hardware Trojans implemented in an IC will leave a footprint at the active layer. In this paper, we propose a new technique based on rapid backside SEM imaging and advanced computer vision algorithms to detect any subtle changes at the active region of transistors that can show the existence of a hardware Trojan. Here, we are only concerned with untrusted foundry problem, where it is assumed the attacker has access to a golden layout/image of the IC. This is a common threat model for those organizations that fully design their IC but need access to untrusted foundry for fabrication. SEM image from a backside thinned golden IC is compared with a low-quality SEM image of an IC under authentication (IUA). We perform image processing to both golden IC and IUA images to remove noise. We have developed a computer vision-based framework to detect hardware Trojans based on their structural similarity. The results demonstrate that our technique is quite effective at detecting Trojans and significantly faster than full chip reverse engineering. One of the major advantages of our technique is that it does not rely on the functionality of the circuit, rather the real physical structure to detect malicious changes performed by the untrusted foundry.

Method taking into account process dispersion to detect hardware Trojan Horse by side-channel analysis

Abstract: Hardware trojans inserted in integrated circuits have received special attention of researchers. Most of the recent researches focus on detecting the presence of hardware trojans through various techniques like reverse engineering, test/verification methods and side-channel analysis (SCA). Previous works using SCA for trojan detection are based on power measurements, or even simulations. When using real silicon, the results are strongly biased by the process variations, the exact size of the trojan, and its location. In this paper, we propose a metric to measure the impact of these parameters. For the first time, we give the detection probability of a trojan as a function of its activity, even if untriggered. Moreover, we use electromagnetic field as side-channel, as it provides a better spatial and temporal resolution than power measurements. We conduct a proof of concept study using an AES-128 cryptographic core running on a set of 10 Virtex-5 FPGA. Our results show that, using this metric, there is a probability superior than 99 % with a false negative rate of 0.017 % to detect a HT bigger than 1 % of the original circuit.

Neural Trojans

Abstract: While neural networks demonstrate stronger capabilities in pattern recognition nowadays, they are also becoming larger and deeper As a result, the effort needed to train a network also increases dramatically In many cases, it is more practical to use a neural network intellectual property (IP) that an IP vendor has already trained As we do not know about the training process, there can be security threats in the neural IP: the IP vendor (attacker) may embed hidden malicious functionality, ie neural Trojans, into the neural IP We show that this is an effective attack and provide three mitigation techniques: input anomaly detection, re-training, and input preprocessing All the techniques are proven effective The input anomaly detection approach is able to detect 998% of Trojan triggers although with 122% false positive The re-training approach is able to prevent 941% of Trojan triggers from triggering the Trojan although it requires that the neural IP be reconfigurable In the input preprocessing approach, 902% of Trojan triggers are rendered ineffective and no assumption about the neural IP is needed