PointCAT: Contrastive Adversarial Training for Robust Point Cloud Recognition.

Notwithstanding the prominent performance shown in various applications, point cloud recognition models have often suffered from natural corruptions and adversarial perturbations. In this paper, we delve into boosting the general robustness of point cloud recognition, proposing Point-Cloud Contrastive Adversarial Training (PointCAT). The main intuition of PointCAT is encouraging the target recognition model to narrow the decision gap between clean point clouds and corrupted point clouds by devising feature-level constraints rather than logit-level constraints. Specifically, we leverage a supervised contrastive loss to facilitate the alignment and the uniformity of hypersphere representations, and design a pair of centralizing losses with dynamic prototype guidance to prevent features from deviating outside their belonging category clusters. To generate more challenging corrupted point clouds, we adversarially train a noise generator concurrently with the recognition model from the scratch. This differs from previous adversarial training methods that utilized gradient-based attacks as the inner loop. Comprehensive experiments show that the proposed PointCAT outperforms the baseline methods, significantly enhancing the robustness of diverse point cloud recognition models under various corruptions, including isotropic point noises, the LiDAR simulated noises, random point dropping, and adversarial perturbations. Our code is available at: https://github.com/shikiw/PointCAT.

Modification of ultrasound-pretreated montmorillonite using poly(diallyldimethylammonium chloride) for W and Mo separation and the sequential application in removal of heavy metals.

The use of a resin to selectively separate thiomolybdate from a tungsten (W) feed solution is a well-known protocol for achieve high-purity W products; however, the regeneration of saturated resin is laborious. In this study, poly(diallyl dimethyl ammonium chloride) (PDADMA) was used to modify ultrasound-pretreated montmorillonite (Mt) for W and molybdenum (Mo) separation for the first time, and the resultant tetrathiomolybdate (MoS42-)-loaded composite was further tested to remove heavy metals instead of regeneration. Among the three variables of ultrasound pretreatment, that is, Mt concentration, ultrasound power, and treatment time, the Mt concentration exhibited the most significant influence followed by ultrasound power on the separation performance of W and Mo. Compared to the distance of the interlayer space and the surface charge of the modified Mt, the PDADMA content showed a closer correlation with the W/Mo separation coefficient. Assisted by Box-Behnken design, with Mt concentration of 6.9 g/L, ultrasound power of 593.8 W, and treatment time of 13.8 min, the composite with the greatest separation coefficient was obtained. The adsorption of Cu(II) on the optimal W/Mo separation-derived composite was ascribed to the formation of Cu-S complexes, while that of Pb(II) was attributed to complexation and surface precipitation. In contrast, ion exchange with the initially loaded anions, reduction by sulfide to Cr(III), and formation of Cr(III)-S complexes accounted for Cr(VI) removal. The adsorption of Cu(II) and Pb(II) equilibrated faster and showed higher acid-resistance than that of Cr(VI). The adsorption capacities for Cu(II), Pb(II), and Cr(VI) were 0.535, 1.398, and 0.882 mmol/g, respectively. Applying PDADMA to modify Mt as a reagent for W/Mo separation was feasible, and the derived composite was capable of removing cationic and anionic heavy metals.

Poison Ink: Robust and Invisible Backdoor Attack.

Recent research shows deep neural networks are vulnerable to different types of attacks, such as adversarial attacks, data poisoning attacks, and backdoor attacks. Among them, backdoor attacks are the most cunning and can occur in almost every stage of the deep learning pipeline. Backdoor attacks have attracted lots of interest from both academia and industry. However, most existing backdoor attack methods are visible or fragile to some effortless pre-processing such as common data transformations. To address these limitations, we propose a robust and invisible backdoor attack called "Poison Ink". Concretely, we first leverage the image structures as target poisoning areas and fill them with poison ink (information) to generate the trigger pattern. As the image structure can keep its semantic meaning during the data transformation, such a trigger pattern is inherently robust to data transformations. Then we leverage a deep injection network to embed such input-aware trigger pattern into the cover image to achieve stealthiness. Compared to existing popular backdoor attack methods, Poison Ink outperforms both in stealthiness and robustness. Through extensive experiments, we demonstrate that Poison Ink is not only general to different datasets and network architectures but also flexible for different attack scenarios. Besides, it also has very strong resistance against many state-of-the-art defense techniques.

Immobilization of W(VI) and/or Cr(VI) in soil treated with montmorillonite modified by a gemini surfactant and tetrachloroferrate (FeCl4-).

The coexistence of highly toxic chromium (Cr) and the emerging contaminant tungsten (W) in the soil adjacent to W mining areas is identified. Immobilization of W and/or Cr is vital for the safe utilization of contaminated soil. In this study, the cationic gemini surfactant (butane-1,4-bis(dodecyl dimethyl ammonium bromide)) and tetrachloroferrate (FeCl4-)-modified montmorillonite (FeOMt) was applied to investigate the retention performance of W and/or Cr in the soil. Regardless of the initially spiked amount of WO42- and/or CrO42-, the W and/or Cr leached in soil solution was rapidly immobilized within 5 min. The immobilization rates of W and/or Cr in the single and binary soil systems were stably maintained against the variations in pH and coexisting anion. FeOMt showed more favorable performance in the retention of W and/or Cr with respect to the precursors (i.e., the original Mt and surfactant-modified Mt) and efficiently inhibited the phytotoxicity and bioaccumulation of W and/or Cr in mung beans. Due to the ion exchange, complexation, reduction, and flocculation, the addition of FeOMt transformed W and/or Cr from exchangeable/carbonate species to reducible/oxidizable fractions, reducing the environmental risk. FeCl4- complex, as a byproduct of the steel pickling process in industry, plays the pivotal role in the efficient retention of W and Cr. Based on the facile synthesis procedure and the efficient performance, the use of FeOMt for the amendment of W- and/or Cr-contaminated soil is feasible and promising.

Gemini surfactant-modified montmorillonite with tetrachloroferrate (FeCl4-) as a counterion simultaneously sequesters nitrate and phosphate from aqueous solution.

Alkyl quaternary ammonium-modified clay minerals, which are common environmentally friendly materials, have been widely studied and applied for the removal of pollutants. However, there are few reports on functionalizing the counterions to expand the application. In this study, the cationic gemini surfactant butane-1,4-bis(dodecyl dimethyl ammonium bromide) (gBDDA) and tetrachloroferrate (FeCl4-) are designed to modify montmorillonite (Mt), and the obtained FeCl4-/Gemini-Mt composite (FeOMt) is used for the removal of nitrate and/or phosphate from aqueous solution. The successful intercalation of gBDDA and favorable loading of FeCl4- into FeOMt are suggested by the characterization results of X-ray diffraction and Raman spectroscopy. Nitrate and/or phosphate are rapidly sequestered, and the respective maximum uptakes of 8.77 (N) and 28.1 (P) mg/g in the binary system are obtained. The phosphate uptake is stably maintained against many coexisting ions, but the nitrate uptake decreases with the increase in ionic strength. FeOMt is reusable and shows comparable uptake for nitrate and phosphate with respect to gBDDA-modified Mt and polymerized ferric chloride. Considering the multi-functionality and facile synthesis, FeOMt shows promising potential in the purification of wastewater contaminated simultaneously by poorly hydrated anions (e.g., ClO4-, TcO4-, etc.) and iron-selective anions (e.g., H2AsO4-, etc.).

Synergistic effect of ClO4- and Sr2+ adsorption on alginate-encapsulated organo-montmorillonite beads: Implication for radionuclide immobilization.

Perchlorate (ClO4-) and pertechnetate (TcO4-) exhibit similar adsorption characteristics on alkyl quaternary ammonium-modified montmorillonite (Mt), and 99mTcO4- normally coexists with 90Sr2+ in radionuclide-contaminated water. In this study, hexadecyl pyridinium (HDPy)-modified Mt (OMt) was encapsulated in alginate beads to inhibit HDPy release and simultaneously immobilize ClO4- and Sr2+ ions. The release of HDPy was remarkably reduced (78 times) from OMt after alginate encapsulation. Adsorption of ClO4- and Sr2+ on the obtained composite demonstrated synergistic effects, with adsorption capacities reaching 0.542 and 0.484 mmol/g, respectively. Compared to the single-adsorbate system, adsorption capacities of ClO4- and Sr2+ increased significantly. The characterization of solids using X-ray diffraction, Fourier transform infrared spectroscopy, 13C nuclear magnetic resonance, and X-ray photoelectron spectroscopy, as well as the chemical analysis of the aqueous solution, demonstrated that HDPy+-COO- disintegration accounted for the adsorption synergy. HDPy was extracted from the Mt interlayer space during the synthesis of OMt/alginate and then partially re-intercalated back after interacting with ClO4- during the adsorption of ClO4- and/or Sr2+. In the binary-adsorbate system, the synergy-induced adsorption capacity was superior to many previously reported adsorbents, implying that OMt/alginate beads can be a promising adsorbent for the remediation of aqueous solutions contaminated with multiple radionuclides.