Deep learning methods for scientific and industrial research
by G.K. Patra, Kantha Rao Bhimala, Ashapurna Marndi, Saikat Chowdhury, Jarjish Rahaman, Sutanu Nandi, Ram Rup Sarkar, K.C. Gouda, K.V. Ramesh, Rajesh P. Barnwal, Siddhartha Raj, Anil Saini
Abstract
Deep learning (DL) is a very powerful computational tool for various applications in scientific and industrial research which can be real-time implemented for societal benefits. Several factors impact the development of optimized DL models for better prediction including the amount of quality sample data, domain-specific knowledge, and the architecture of the model for extraction of the useful features/patterns from the data. The present chapter demonstrates the state-of-the-art DL methodologies used by the researchers from different laboratories under the Council of Scientific and Industrial Research (CSIR), India to solve important research activities across several sectors like Medical, Healthcare, Agriculture, Energy, etc. The Convolutional Neural Network (CNN) techniques are utilized for Tumor diagnosis, classifying molecular subtypes of glioma tissues, and predicting driver gene mutations in glioma. Similarly, the Long short-term memory (LSTM) model is applied for the assessment of crop production, and transfer learning is used for the classification of tea leaves. Further, the ensemble LSTM methodology is implemented for short-term prediction of wind speed to enhance the renewable energy sectors. Finally, the multivariate LSTM models were developed by integrating the weather parameters for the prediction of covid-19 spread over different states in India which is an input for policy planning and supply chain management during the pandemic time. All the use cases are being validated and the results are quite satisfying and provide confidence for the real-time application of DL for scientific and industrial research and societal benefit to the common people.
Source: https://doi.org/10.1016/bs.host.2022.12.002
Decoding low-frequency climate variations: A case study on ENSO and ocean surface warming
by Rameshan Kallummal
Current perspectives on lower frequency variations and secular warming have predominantly been shaped by traditional anomalies that assume an annual cycle (AC) with a time-invariant amplitude. However, this anomaly framework falls short in capturing the complexity of multiple periodic modes with intricate waveforms and time dependent amplitude—the traits, in general, shared by externally forced responses of complex dynamical systems. By allowing interannual amplitude modulation of ACs, we show that the monotonic amplitude increases of the first AC of Sea Surface Temperature (SST) are manifested as the basin-wide secular warming of the ocean surface. Notably, the first SST-AC exhibits significant interannual variances and the largest linear warming rates in the Pacific Warm Pool. While the linear warming pattern mirrors that of a long-term mean SST, it depicts an entirely different warming pattern on the surface of the tropical Pacific Ocean compared to those reported so far. Moreover, all interannual warm (El Niño) and cold (La Niña) events in tropical Pacific regions are abnormal interannual modulations in the third and fourth ACs of SST, respectively. Specifically, a strong El Niño event occurs when a positive amplitude modulation leads to the overlap of two consecutive positive phases of the third AC of SST. Conversely, the absence of such overlaps during negative amplitude modulations significantly contributes to the positive skewness of SST anomalies. No systematic decadal changes in the zonal propagation characteristics of SST in the eastern Pacific (EP) and central Pacific (CP) regions were detected. These findings underscore that the secular warming and low-frequency events in EP and CP are intrinsic to three distinct ACs.
Data sequence signal manipulation in multipath TCP (MPTCP): The vulnerability, attack and its detection
by V Anil Kumar and Debabrata Das
Multipath Transmission Control Protocol (MPTCP) is an innovative next-generation transport protocol standardized by the Internet Engineering Task Force (IETF) to overcome the single path limitation of the Transmission Control Protocol (TCP). MPTCP augments TCP with a new set of signaling options for seamless transmission and reception of application data across multiple interlinked TCP connections called subflows. In this paper, we focus on a new security concern associated with the signal exchanging process of MPTCP. To the best of our knowledge, for the first time, this paper exposes that MPTCP signal exchange scheme is vulnerable to a sophisticated packet spoofing technique, which we name as Data Sequence Signal (DSS) manipulation. We implement the vulnerability, create attack scenarios in Linux Kernel and conduct experiments over emulated testbed to demonstrate the existence of the vulnerability and means of exploiting it for powerful attacks. Our results show that DSS manipulation can be tactically exploited, on top of TCP optimistic ACKing, to generate non-responsive traffic like Denial-of-Service (DoS) attack flood. Particularly, we demonstrate two new adverse scenarios, where a MPTCP sender is forced to: (a) transmit at a rate significantly higher than the bottleneck link bandwidth, and (b) induce high intensity and harmful packet bursts at line-rate called Maliciously-induced-Bursts (MiBs). We also show that the non-responsive traffic resulting from the attack can suppress genuine congestion controlled traffic to the extent of causing DoS attack. We capture and analyze the dynamics of important MPTCP parameters, like send buffer occupancy of meta and subflow sockets, congestion window and flightsize to highlight the attack impact. DSS manipulation originates from a fundamental protocol design limitation rather than from any implementation flaw. We also propose a novel technique called data sequence map skipping for detection and countermeasure against DSS manipulation based attacks.
Crustal deformation rates in Kashmir valley and adjoining regions from continuous GPS measurements from 2008 to 2019
by Sridevi Jade, Ramees R. Mir, Chiranjeevi G. Vivek, T. S. Shrungeshwara, I. A. Parvez, Rakesh Chandra, D. Suri Babu, S. Vishal Gupta, Ankit, Siva Sai Kumar Rajana & V. K. Gaur
We present GPS velocities in Kashmir valley and adjoining regions from continuous Global Positioning System (cGPS) network during 2008 to 2019. Results indicate total arc normal shortening rates of ~ 14 mm/year across this transect of Himalaya that is comparable to the rates of ~ 10 to 20 mm/year reported else-where in the 2500 km Himalaya Arc. For the first time in Himalayas, arc-parallel extension rate of ~ 7 mm/year was recorded in the Kashmir valley, pointing to oblique deformation. Inverse modeling of the contemporary deformation rates in Kashmir valley indicate oblique slip of ~ 16 mm/year along the decollement with locking depth of ~ 15 km and width of ~ 145 km. This result is consistent with the recorded micro-seismicity and low velocity layer at a depth of 12 to 16 km beneath the Kashmir valley obtained from collocated broadband seismic network. Geodetic strain rates are consistent with the dislocation model and micro-seismic activity, with high strain accumulation (~ 7e−08 maximum compression) to the north of Kashmir valley and south of Zanskar ranges. Assuming the stored energy was fully released during 1555 earthquake, high geodetic strain rate since then and observed micro-seismicity point to probable future large earthquakes of Mw ~ 7.7 in Kashmir seismic gap.
Page 25 of 33
