scholarly journals Attribution of genetic engineering: A practical and accurate machine-learning toolkit for biosecurity

2020 ◽  
Author(s):  
Ethan C. Alley ◽  
Miles Turpin ◽  
Andrew Bo Liu ◽  
Taylor Kulp-McDowall ◽  
Jacob Swett ◽  
...  
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Genetic Engineering ◽  
Recurrent Neural Networks ◽  
International Security ◽  
Responsible Innovation ◽  
Nation State ◽  
Dna Motifs ◽  
Simple Models

AbstractThe promise of biotechnology is tempered by its potential for accidental or deliberate misuse. Reliably identifying telltale signatures characteristic to different genetic designers, termed genetic engineering attribution, would deter misuse, yet is still considered unsolved. Here, we show that recurrent neural networks trained on DNA motifs and basic phenotype can reach 70% attribution accuracy distinguishing between over 1,300 labs. To make these models usable in practice, we introduce a framework for weighing predictions against other investigative evidence using calibration, and bring our model to within 1.6% of perfect calibration. Additionally, we demonstrate that simple models can accurately predict both the nation-state-of-origin and ancestor labs, forming the foundation of an integrated attribution toolkit which should promote responsible innovation and international security alike.

scholarly journals A machine learning toolkit for genetic engineering attribution to facilitate biosecurity

2020 ◽  
Vol 11 (1) ◽  
Author(s):  
Ethan C. Alley ◽  
Miles Turpin ◽  
Andrew Bo Liu ◽  
Taylor Kulp-McDowall ◽  
Jacob Swett ◽  
...  
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Genetic Engineering ◽  
Recurrent Neural Networks ◽  
International Security ◽  
Responsible Innovation ◽  
Nation State ◽  
Dna Motifs ◽  
Phenotype Data ◽  
Simple Models

AbstractThe promise of biotechnology is tempered by its potential for accidental or deliberate misuse. Reliably identifying telltale signatures characteristic to different genetic designers, termed ‘genetic engineering attribution’, would deter misuse, yet is still considered unsolved. Here, we show that recurrent neural networks trained on DNA motifs and basic phenotype data can reach 70% attribution accuracy in distinguishing between over 1,300 labs. To make these models usable in practice, we introduce a framework for weighing predictions against other investigative evidence using calibration, and bring our model to within 1.6% of perfect calibration. Additionally, we demonstrate that simple models can accurately predict both the nation-state-of-origin and ancestor labs, forming the foundation of an integrated attribution toolkit which should promote responsible innovation and international security alike.

Machine Learning-Based Malware Detection using Recurrent Neural Networks

2019 ◽  
Author(s):  
Miguel Oyler-Castrillo ◽  
Nicolas Bohm Agostini ◽  
Gadiel Sznaier ◽  
David Kaeli
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Recurrent Neural Networks ◽  
Malware Detection

scholarly journals Train Delay Prediction using Machine Learning

2019 ◽  
Vol 9 (2) ◽  
pp. 1312-1315
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Recurrent Neural Networks ◽  
Weather Conditions ◽  
Severe Weather ◽  
Average Error ◽  
Machine Learning Method ◽  
Long Distance ◽  
Passenger Trains ◽  
Delay Prediction

Indian Railways operates both long distance and suburban passenger trains and freight services daily in the country. Trains get delayed frequently due to several reasons such as, severe weather conditions such as fog, traffic, signal failure, derailing of trains, accidents, etc, and this delay is propagated from station to station. If we can predict this in advance - it would be of great help for the commuters to plan their journey either for an earlier departure or postpone, and also lets railways to take measures to avoid delays further. In this paper, we used decision tree, a machine learning method used for predicting train delays, and Recurrent Neural Networks distinguished with various fixtures. For predicting train delays, Recurrent Neural networks with 2 layers and 22 neurons per each layer gave best results with an average error of 122 seconds

Evaluation of Recurrent Neural Network and its Variants for Intrusion Detection System (IDS)

2017 ◽  
Vol 8 (3) ◽  
pp. 43-63 ◽  
Author(s):  
R Vinayakumar ◽  
K.P. Soman ◽  
Prabaharan Poornachandran
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Intrusion Detection ◽  
Recurrent Neural Networks ◽  
Machine Learning Algorithms ◽  
Time Range ◽  
Sequence Information ◽  
Time Lags ◽  
Data Set ◽  
Arbitrary Length

This article describes how sequential data modeling is a relevant task in Cybersecurity. Sequences are attributed temporal characteristics either explicitly or implicitly. Recurrent neural networks (RNNs) are a subset of artificial neural networks (ANNs) which have appeared as a powerful, principle approach to learn dynamic temporal behaviors in an arbitrary length of large-scale sequence data. Furthermore, stacked recurrent neural networks (S-RNNs) have the potential to learn complex temporal behaviors quickly, including sparse representations. To leverage this, the authors model network traffic as a time series, particularly transmission control protocol / internet protocol (TCP/IP) packets in a predefined time range with a supervised learning method, using millions of known good and bad network connections. To find out the best architecture, the authors complete a comprehensive review of various RNN architectures with its network parameters and network structures. Ideally, as a test bed, they use the existing benchmark Defense Advanced Research Projects Agency / Knowledge Discovery and Data Mining (DARPA) / (KDD) Cup ‘99' intrusion detection (ID) contest data set to show the efficacy of these various RNN architectures. All the experiments of deep learning architectures are run up to 1000 epochs with a learning rate in the range [0.01-0.5] on a GPU-enabled TensorFlow and experiments of traditional machine learning algorithms are done using Scikit-learn. Experiments of families of RNN architecture achieved a low false positive rate in comparison to the traditional machine learning classifiers. The primary reason is that RNN architectures are able to store information for long-term dependencies over time-lags and to adjust with successive connection sequence information. In addition, the effectiveness of RNN architectures are shown for the UNSW-NB15 data set.

scholarly journals Prediction of ISO 9001:2015 Audit Reports According to its Major Clauses using Recurrent Neural Networks

2019 ◽  
Vol 8 (2) ◽  
pp. 1773-1778
Keyword(s):  
Machine Learning ◽  
Data Mining ◽  
Neural Networks ◽  
Quality Assurance ◽  
Recurrent Neural Networks ◽  
Management System ◽  
Iso 9001 ◽  
Top Management ◽  
Digital Documents ◽  
Bidirectional Lstm

The Quality Assurance Department of the educational sectors is rapidly generating digital documents. The continuous increase of digital documents may become a risk and challenge in the future. Interpreting and analyzing those digital data in a short period of time is very critical and crucial for the top management to support their decisions. By this purpose, this paper explored the possibility of machine learning and data mining process to improve the Quality Assurance Management System process, specifically in the Quality Audit procedures and generation of management reports. The researchers developed a machine learning model that predicts an audit report according to the major clauses of the ISO 9001:2015 Quality Management System (QMS) Requirements. The proposed data mining process helps the top management to identify which principles of the ISO 9001:2015 QMS Requirements they are lacking. The authors used four different Recurrent Neural Networks (RNNs) as a classifier; (1) Long Short Term-Memory (LSTM), (2) Bidirectional-LSTM, (3) Deep-LSTM and a (4) Deep-Bidirectional-LSTM Recurrent Neural Networks with a combine word representation models (word encoding plus an embedding dimension layer). The Deep-Bidirectional-LSTM outperformed the other three RNN models. Where it achieved an average classification accuracy of 91.10%

scholarly journals A Survey on Performing E-Voting through Facial Recognition

2021 ◽  
pp. 141-147
Author(s):  
Pratik Hopal ◽  
Alkesh Kothar ◽  
Swamini Pimpale ◽  
Pratiksha More ◽  
Jaydeep Patil
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Recurrent Neural Networks ◽  
Facial Recognition ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Voting Systems ◽  
Human Errors ◽  
Research Article ◽  
The Future

The election procedure is one of the most essential processes to take place in a democracy. Even though there have been immense technological advancements, the process of election has been highly limited. Most of the election procedures have been performed using ballot boxes which is an old process and needs to be updated. The security of such practices is also a concern as the identification of the voters is being done manually by the election officers. This process also needs an improvement to increase accuracy and reduce human errors by automating the process. Therefore, for this purpose, this research article analyzes the previous researches on this paradigm. This allows an effective understanding of the machine learning algorithms that are used for automatic facial recognition in the E-voting systems. This paper comes to the conclusion that the Recurrent Neural Networks are best suited for such an application for facial recognition. The future editions of this research will elaborate more on the proposed system in detail.

Sign in / Sign up

Export Citation Format

Share Document