Search

Deep neural networks (DNNs) are vulnerable to adversarial examples, which are crafted by adding small, human-imperceptible perturbations to the original images, but make the model output inaccurate predictions. Before DNNs are deployed, adversarial attacks can thus be an important method to evaluate and select robust models in safety-critical applications. However, under the challenging black-box setting, the attack success rate, i.e., the transferability of adversarial examples, still needs to be improved. Based on image augmentation methods, this paper found that random transformation of image brightness can eliminate overfitting in the generation of adversarial examples and improve their transferability. In light of this phenomenon, this paper proposes an adversarial example generation method, which can be integrated with Fast Gradient Sign Method (FGSM)-related methods to build a more robust gradient-based attack and to generate adversarial examples with better transferability. Extensive experiments on the ImageNet dataset have demonstrated the effectiveness of the aforementioned method. Whether on normally or adversarially trained networks, our method has a higher success rate for black-box attacks than other attack methods based on data augmentation. It is hoped that this method can help evaluate and improve the robustness of models. [ABSTRACT FROM AUTHOR]

Aiming at the safety problems of trusted computing platform due to the Apps lack of trust, this paper presented a software identification distributing system, and introduced its function modules and workflow. Software identification acted as its safety carriers and authentication means, software behavior statement solved the problem of legitimacy and trust before distribution. It designed an authentication protocol to ensure the integrity of software during distribution, and provided a recommended strategies for choosing suitable software based on trustworthy behavior. Application examples demonstrate that the system has a higher research and practical value of enhancing the trustworthiness of the platform. [ABSTRACT FROM AUTHOR]

This paper designed a kind of active measurement model named mixed-chain & character pre-setting active-measurement model(MC-CPSAM). By making improvement of the traditional chain structure of trust, MC-CPSAM made the metric agents more trusted. It made an achievement for a kind of active measurement by protecting metric agents and upgrading rights of metric agents. At last, it designed a hardware structure based on MC-CPSAM and made an attestation for the security of this model. [ABSTRACT FROM AUTHOR]

Background: Epidemiological surveys and studies with animal models have established a relationship between maternal stress and affective disorders in their offspring. However, whether maternal depression before pregnancy influences behaviour and related neurobiological mechanisms in the offspring has not been studied.Results: A social defeat stress (SDS) maternal rat model was established using the resident-intruder paradigm with female specific pathogen-free Wistar rats and evaluated with behavioural tests. SDS maternal rats showed a significant reduction in sucrose preference and locomotor and exploratory activities after 4 weeks of stress. In the third week of the experiment, a reduction in body weight gain was observed in SDS animals. Sucrose preference, open field, the elevated-plus maze, light-dark box, object recognition, the Morris water maze, and forced swimming tests were performed using the 2-month-old male offspring of the female SDS rats. Offspring subjected to pre-gestational SDS displayed enhanced anxiety-like behaviours, reduced exploratory behaviours, reduced sucrose preference, and atypical despair behaviours. With regard to cognition, the offspring showed significant impairments in the retention phase of the object recognition test, but no effect was observed in the acquisition phase. These animals also showed impairments in recognition memory, as the discrimination index in the Morris water maze test in this group was significantly lower for both 1 h and 24 h memory retention compared to controls. Corticosterone, adrenocorticotropic hormone, and monoamine neurotransmitters levels were determined using enzyme immunoassays or radioimmunoassays in plasma, hypothalamus, left hippocampus, and left prefrontal cortex samples from the offspring of the SDS rats. These markers of hypothalamic-pituitary-adrenal axis responsiveness and the monoaminergic system were significantly altered in pre-gestationally stressed offspring. Brain-derived neurotrophic factor (BDNF), cyclic adenosine monophosphate response element binding protein (CREB), phosphorylated CREB (pCREB), and serotonin transporter (SERT) protein levels were evaluated using western blotting with right hippocampus and right prefrontal cortex samples. Expression levels of BDNF, pCREB, and SERT in the offspring were also altered in the hippocampus and in the prefrontal cortex; however, there was no effect on CREB.Conclusion: We conclude that SDS before pregnancy might induce depressive-like behaviours, cognitive deficits, and neurobiological alterations in the offspring. [ABSTRACT FROM AUTHOR]