Deep neural networks are susceptible to poisoning attacks by purposely polluted training data with specific triggers. As existing episodes mainly focused on attack success rate with patch-based samples, defense algorithms can easily detect these poisoning samples. We propose DeepPoison, a novel adversarial network of one generator and two discriminators, to address this problem. Specifically, the generator automatically extracts the target class' hidden features and embeds them into benign training samples. One discriminator controls the ratio of the poisoning perturbation. The other discriminator works as the target model to testify the poisoning effects. The novelty of DeepPoison lies in that the generated poisoned training samples are indistinguishable from the benign ones by both defensive methods and manual visual inspection, and even benign test samples can achieve the attack. Extensive experiments have shown that DeepPoison can achieve a state-of-the-art attack success rate, as high as 91.74%, with only 7% poisoned samples on publicly available datasets LFW and CASIA. Furthermore, we have experimented with high-performance defense algorithms such as autodecoder defense and DBSCAN cluster detection and showed the resilience of DeepPoison.

No abstract

This study investigated the effects of addition of inorganic and organic polymer flocculants as filter aids on mitigation of membrane fouling of membrane bioreactors (MBRs). A series of dead-end filtration batch tests operated at constant pressure were conducted to screen effective filter aid candidates. Continuous MBR operation was carried out subsequently to identify the optimal filter aids selected from the batch tests. Various biomass parameters such as capillary suction time, zeta (ζ) potential, floc size, extracellular polymer substances, and soluble microbial products (SMP) in different MBRs were investigated to study the mechanism leading to the flux improvement. The experimental results demonstrated the significant transmembrane pressure reduction through dosing the selected filter aids. Regular addition of filter aids could extend the stable filtration performance of mixed liquor. Filter aids could mitigate membrane fouling mainly through reducing the concentration of SMP of the mixed liquor. Besides, inorganic additives improved total phosphorus removal of MBR effluent, while organic additives enhanced the sludge dewatering performance through forming fairly large sludge floc size and higher ζ potential. It was also observed that adding filter aids had no negative impact on both chemical oxygen demand and ammonium removal of the MBR system.

No abstract

Fire and explosion accidents and reduced energy utilization due to poor cycling stability of lithium-ion batteries (LIBs) caused by inevitable internal temperature rise during high-rate operations have become a growing concern. Herein, a dual-functional carbon nanotube/hygroscopic salt (DFCNT/ HS) film with effective passive cooling performance and fire insulation for the safe usage of practical LIBs under extremely fast discharging conditions is reported. The DFCNT/HS film based on the cooling mechanism of self-adaptive moisture absorption/desorption delivers a high cooling power of 32.9 W m −2 K −1 , which can reduce the maximum temperature of a 18650-3.6 V/2.0 Ah LIB by 11.2 and 17.4 °C at discharging rates of 10 and 15 C, respectively. Covering the cooling film, the battery discharges 23.6 Ah more total capacity at 10 within 500 cycles. What is challenging, almost three-fold extended lifetime of 425 cycles is achieved at 15 C with an extra total capacity of 467.2 Ah. Meanwhile, the developed film also shows an excellent hightemperature resistance up to 540 °C, which can alleviate the devastating fire propagation. The fast heat dissipation and excellent fire insulation as well as the mechanical flexibility and manufacturing scalability make this new material promising for safe usage of high-rate LIBs with zero energy consumption.