Deep learning (DL) models are vulnerable to adversarial attacks, by adding a subtle perturbation which is imperceptible to the human eye, a convolutional neural network (CNN) can lead to erroneous results, which greatly reduces the reliability and security of the DL tasks. Considering the wide application of modulation recognition in the communication field and the rapid development of DL, by adding a well-designed adversarial perturbation to the input signal, this article explores the performance of attack methods on modulation recognition, measures the effectiveness of adversarial attacks on signals, and provides the empirical evaluation of the reliabilities of CNNs. The results indicate that the accuracy of the target model reduce significantly by adversarial attacks, when the perturbation factor is 0.001, the accuracy of the model could drop by about 50% on average. Among them, iterative methods show greater attack performances than that of one-step method. In addition, the consistency of the waveform before and after the perturbation is examined, to consider whether the added adversarial examples are small enough (i.e., hard to distinguish by human eyes). This article also aims at inspiring researchers to further promote the CNNs reliabilities against adversarial attacks.

Ablation of nonmuscle myosin (NM) II-B in mice during embryonic development leads to marked enlargement of the cerebral ventricles and destruction of brain tissue, due to hydrocephalus. We have identified a transient mesh-like structure present at the apical border of cells lining the spinal canal of mice during development. This structure, which only contains the II-B isoform of NM, also contains β-catenin and N-cadherin, consistent with a role in cell adhesion. Ablation of NM II-B or replacement of NM II-B with decreased amounts of a mutant (R709C), motor-impaired NM II-B in mice results in collapse of the mesh-like structure and loss of cell adhesion. This permits the underlying neuroepithelial cells to invade the spinal canal and obstruct cerebral spinal fluid flow. These defects in the CNS of NM II-B–ablated mice seem to be the cause of hydrocephalus. Interestingly, the mesh-like structure and patency of the spinal canal can be restored by increasing expression of the motor-impaired NM II-B, which also rescues hydrocephalus. However, the mutant isoform cannot completely rescue neuronal cell migration. These studies show that the scaffolding properties of NM II-B play an important role in cell adhesion, thereby preventing hydrocephalus during mouse brain development.

Ablation of nonmuscle myosin (NM) II-A or NM II-B results in mouse embryonic lethality. Here, we report the results of ablating NM II-C as well as NM II-C/II-B together in mice. NM II-C ablated mice survive to adulthood and show no obvious defects compared with wild-type littermates. However, ablation of NM II-C in mice expressing only 12% of wild-type amounts of NM II-B results in a marked increase in cardiac myocyte hypertrophy compared with the NM II-B hypomorphic mice alone. In addition, these hearts develop interstitial fibrosis associated with diffuse N-cadherin and β-catenin localization at the intercalated discs, where both NM II-B and II-C are normally concentrated. When both NM II-C and II-B are ablated the B⁻C⁻/B⁻C⁻ cardiac myocytes show major defects in karyokinesis. More than 90% of B⁻C⁻/B⁻C⁻ myocytes demonstrate defects in chromatid segregation and mitotic spindle formation accompanied by increased stability of microtubules and abnormal formation of multiple centrosomes. This requirement for NM II in karyokinesis is further demonstrated in the HL-1 cell line derived from mouse atrial myocytes, by using small interfering RNA knockdown of NM II or treatment with the myosin inhibitor blebbistatin. Our study shows that NM II is involved in regulating cardiac myocyte karyokinesis by affecting microtubule dynamics.

Background—: The 3 isoforms of nonmuscle myosin (NM) II (NMII-A, NMII-B, and NMII-C) play various roles during mouse embryonic development. Previous work, using knockout and hypomorphic mice, showed that Myh10 encoding myosin heavy chain II-B is critical for cardiac and brain development. Ablating or decreasing NMII-B by 80% results in cardiac (ventricular septal defect, double outlet of the right ventricle) and brain defects but not midline fusion defects. Neither NMII-A nor II-C seems to play roles in early myocardial development. Methods and Results—: We had previously generated point mutant knock-in mice and now report novel findings as a result of expressing motor-deficient NMII-B at wild-type levels. Homozygous mice die at embryonic day 14.5 in cardiac failure, exhibiting abnormalities not seen in NMII-B null and hypomorphic mice: a failure in midline fusion resulting in a cleft palate, ectopia cordis, and a large omphalocele. Fusion of the sternum and endocardial cushions is impaired in the mutant mice associated with a failure in apoptosis of the mesenchymal cells. Failure to disassemble myocyte cell–cell adhesions during cardiac outflow tract development contributes to impaired outflow tract myocardialization and displacement of the aorta to the right ventricle. Conclusions—: Expression of motor-impaired NMII-B disrupts normal ventral body wall closure because of a dominant-negative effect. This is not because of the loss of NMII-B function but rather a gain-of-function resulting from prolonged cross-linking of NMII-B to actin filaments, thereby interfering with the dynamics of actomyosin cytoskeletal structure. Furthermore, impaired NMII-B motor activity inhibits outflow tract myocardialization, leading to mislocalization of the aorta.