Trojan

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

Cassandra: Detecting Trojaned Networks from Adversarial Perturbations

Abstract: Deep neural networks are being widely deployed for many critical tasks due to their high classification accuracy. In many cases, pre-trained models are sourced from vendors who may have disrupted the training pipeline to insert Trojan behaviors into the models. These malicious behaviors can be triggered at the adversary's will and hence, cause a serious threat to the widespread deployment of deep models. We propose a method to verify if a pre-trained model is Trojaned or benign. Our method captures fingerprints of neural networks in the form of adversarial perturbations learned from the network gradients. Inserting backdoors into a network alters its decision boundaries which are effectively encoded in their adversarial perturbations. We train a two stream network for Trojan detection from its global ($L_\infty$ and $L_2$ bounded) perturbations and the localized region of high energy within each perturbation. The former encodes decision boundaries of the network and latter encodes the unknown trigger shape. We also propose an anomaly detection method to identify the target class in a Trojaned network. Our methods are invariant to the trigger type, trigger size, training data and network architecture. We evaluate our methods on MNIST, NIST-Round0 and NIST-Round1 datasets, with up to 1,000 pre-trained models making this the largest study to date on Trojaned network detection, and achieve over 92\% detection accuracy to set the new state-of-the-art.

Orbital Characteristics of Comets Passing through the 1:1 Commensurability with Jupiter

Abstract: When, in consequence of a close approach, a comet of the Jupiter group changes its osculating semimajor axis from a > 1 to a < 1 (a’ = 1 for Jupiter’s orbit), or vice versa, then the normal case is that of an abrupt change from one side of Jupiter’s orbit to the other one. Under special conditions, however, temporary capture into satellite or ‘Trojan’ status is possible. P/Slaughter-Burnham, the first known comet in temporary 1:1 resonance with Jupiter, sheds some light on the requirements for Trojan captures. In consideration of the recent finding that the Trojan ‘cloud’ around L 4 contains probably as many as 700 Trojans brighter than magnitude 20.9, it is suggested that at least some comets of the Jupiter group may have originated among these accumulations around L 4 and L 5.

Trustworthy computing in a multi-core system using distributed scheduling

Abstract: Hardware Trust is an emerging problem in semiconductor integrated circuit (IC) security due to widespread outsourcing and the stealthy nature of hardware Trojans. Conventional post-manufacturing testing, test generation algorithms and test coverage metrics cannot be readily extended to hardware Trojan detection. As a result there is a need to develop approaches that will ensure trusted in-field operation of ICs, and more generally trust in computing. We present a distributed software scheduling prototype, TADS (Trojan Aware Distributed Scheduling), to achieve a Trojan-activation tolerant trustworthy computing system in a multi-core processor potentially containing hardware Trojans. TADS is designed to be transparent to applications and can run on general purpose multicore PEs without modifications to the operating system or underlying hardware. TADS can, with high confidence, continue to correctly execute its specified queue of job subtasks in the presence of hardware Trojans in the multi-core PEs while learning the individual trustworthiness of the individual PEs. Specially crafted self-checking subtasks called bounty hunters are introduced to accelerate PE trust learning. Also, by learning and maintaining individual PE trustworthiness, the scheduler is able to achieve Trojan containment by scheduling subsequent job subtasks to PEs with high learned trust.

On the accuracy of restricted three-body models for the trojan motion

Abstract: In this note we compare the frequencies of the motion of the Trojan asteroids in the Restricted Three-Body Problem (RTBP), the Elliptic Restricted Three-Body Problem (ERTBP) and the Outer Solar System (OSS) model. The RTBP and ERTBP are well-known academic models for the motion of these asteroids, and the OSS is the standard model used for realistic simulations. Our results are based on a systematic frequency analysis of the motion of these asteroids. The main conclusion is that both the RTBP and ERTBP are not very accurate models for the long-term dynamics, although the level of accuracy strongly depends on the selected asteroid.