Running 10+ models with different arcfhitecture and parameters in anomaly detection use cases on the HMOG dataset with deep feature extratctors and different realistic experiment settings Models include:

• VAE (Variational Auto Encoder)
• b-VAE (Beta Variational Auto Encoder)
• KNN (K-Nearest neighbors)
• ABOD (Angle-based Outlier Detection)
• OCSVM (One-class Support Vector Machine)
• HBOS (Histogram-based Outlier Score)
• PCA (Principal Component Analysis)
• MCD (Minimum Covariance Determinant)
• CBLOF (Cluster-Based Local Outlier Factor)
• LOF (Local Outlier Factor)
• IForest (Isolation Forest)
• FeatureBag (Feature Bagging)
• SCNN (Siamese Convolutional Deep Neural Network) as a deep feature extractor

Part of the following effort: HARNESSING THE POWER OF GENERATIVE MODELS FOR MOBILE CONTINUOUS AND IMPLICIT AUTHENTICATION

Interconnecting the following works:

• Generative AI Models
• Continous Authentication
• Implicit Authentication
• Outlier Detection

Authenticating a user’s identity lies at the heart of securing any information system. A trade off exists currently between user experience and the level of security the system abides by. Using Continuous and Implicit Authentication a user’s identity can be verified without any active participation, hence increasing the level of security, given the continuous verification aspect, as well as the user experience, given its implicit nature. This thesis studies using mobile devices inertial sensors data to identify unique movements and patterns that identify the owner of the device at all times. We implement, and evaluate approaches proposed in related works as well as novel approaches based on a variety of machine learning models, specifically a new kind of Auto Encoder (AE) named Variational Auto Encoder (VAE), relating to the generative models family. We evaluate numerous machine learning models for the anomaly detection or outlier detection case of spotting a malicious user, or an unauthorised entity currently using the smartphone system. We evaluate the results under conditions similar to other works as well as under conditions typically observed in real-world applications. We find that the shallow VAE is the best performer semi-supervised anomaly detector in our evaluations and hence the most suitable for the design proposed. The thesis concludes with recommendations for the enhancement of the system and the research body dedicated to the domain of Continuous and Implicit Authentication for mobile security. Keywords: Machine Learning, Generative Models, Continuous Authentication, Implicit Authentication, Artificial Intelligence

Continauth is based on the work of the Buech in 2019. REF: https://github.com/dynobo/ContinAuth HMOG Dataset is public and can be downloaded from public sources. REF: http://www.cs.wm.edu/~qyang/hmog.html

Home hosts files needed to run this project on Slurm managed resources like the super computers used in Compute Canada, and the Niagara super computer used by us in this work. Continauth hosts src/data and src/utility which have scripts to download/generate data and helper code, respectively. Continauth also hosts notebooks which have all the jupytner nootbooks that have the experiments. Some results that satisfy the size limitations imposed by github are also hosted under the output directory. │

• Clone the repo
• If using a slurm environment run home/request_compute_node_for_24h to get a computing node.
• If using Ijupyter run home/start_joop to start the server
• If needed use an shh tunnel to forward the ports and use foxyproxy to allow your browser to utilize the SOCKS tunnel created
• Start the data processing functions for HMOG dataset by running bash python -m src.data.make_dataset Thanks to dynobo's work the preprocessing is packaged nicely . But it can also run manually

• python -m src.data.unzip_hmog_dataset - Unzip into '/data/external' and optionally remove zip file.
• python -m src.data.transform_to_hdf - Reads CSVs, joins sensor data by time index and store it in HDF format in '/data/processed/' with table key sensors_100hz
• python -m src.data.resample_dataset - Reads data from HDF, resamples it to 25Hz and stores it as separate table sensors_25hz in the same HDF.

• The notebooks under notebooks ContinAuth/notebooks each try a different set of experiments, experiment settings are disclosed in the first section of each notebook, as well as each output result name, which can be found under notebooks/output/chapter-6-1-4-siamese-cnn