Implicit authentication (IA) is a technique that allows the
smart device
A smart device is an electronic device, generally connected to other devices or networks via different wireless protocols (such as Bluetooth, Zigbee, near-field communication, Wi-Fi, LiFi, or 5G) that can operate to some extent interactively and ...
to recognize its owner by being acquainted with his/her behaviors. It is a technique that uses
machine learning
Machine learning (ML) is a field of inquiry devoted to understanding and building methods that 'learn', that is, methods that leverage data to improve performance on some set of tasks. It is seen as a part of artificial intelligence.
Machine ...
algorithms to learn user behavior through various sensors on the smart devices and achieve user identification. Most of the current authentication techniques, e.g., password, pattern lock, finger print and iris recognition, are explicit authentication which require user input. Comparing with explicit authentication, IA is transparent to users during the usage, and it significantly increases the usability by reducing time users spending on login, in which users find it more annoying than lack of cellular coverage.
Model
In Implicit authentication (IA), user behaviors (raw) data are captured by various
sensor
A sensor is a device that produces an output signal for the purpose of sensing a physical phenomenon.
In the broadest definition, a sensor is a device, module, machine, or subsystem that detects events or changes in its environment and sends ...
s embedded in the smart device, and stored in the database preparing for further processing. After filtering out noise and selecting suitable features, the data will be sent to machine learning tool(s) which will train and return a fine-tuned model back to smart device. The smart device then uses the model as signature to identify the current user. Due to the battery and computation limitation of smart device, the training phase, in which most of the computations are carried out, is usually implemented in the remote server. Some lightweight algorithms, e.g.,
Kl divergence
KL, kL, kl, or kl. may refer to:
Businesses and organizations
* KLM, a Dutch airline (IATA airline designator KL)
* Koninklijke Landmacht, the Royal Netherlands Army
* Kvenna Listin ("Women's List"), a political party in Iceland
* KL FM, a Mala ...
, are implemented in the local device as parts of real-time authentication units which control lock mechanism of the device.
The developing of IA model largely depends on the operating systems, which usually adopt Android and iOS, and there are two different approaches to establish IA model, which are device-centric and application-centric. Device-centric approaches, as the traditional way to establish IA model, leverage most of the information gathered by operating system from various sensors, and IA model is directly running above the operating system. Application-centric approaches however achieve IA through establishing individual framework in each app, which executes independently in the
sandbox, and it preserves the intrinsic structure of operating system, while simplifies IA developing.
History
In 1977,
Helen M. Wood indicated that there were two types of bio-metric authentication approaches - physiological and behavioral bio-metrics. The second approach related to user's gait, location information and keystroke patterns. The utilization of the bio-metrics for user authentication had been developed in the field such as: location-based access control, notably
keystroke dynamics
Keystroke dynamics, keystroke biometrics, typing dynamics and typing biometrics refer to the detailed timing information that describes when each key was pressed and released as a person is typing on a computer keyboard.
Science
The behavioura ...
and typing pattern. In 2010, Shi et al. had migrated bio-metrics authentication approach to mobile device which contained many sensors, and significantly increased the accuracy of the authentication, and they called the new approach "implicit authentication".
Due to the fast growth of smart technology, smart device became more and more sophisticated with computational power grew in each year, and it provided the foundation for IA to achieve high accurate and user-friendly authentication. The current IA approaches mainly focused on touch sensor, GPS and accelerometer, and the corresponding techniques were
SVM,
kNN,
GMM and
topic model
In statistics and natural language processing, a topic model is a type of statistical model for discovering the abstract "topics" that occur in a collection of documents. Topic modeling is a frequently used text-mining tool for discovery of hidden ...
.
References
{{Authority control
Authentication methods