RLAuth: A Risk-Based Authentication System Using Reinforcement Learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Conventional authentication systems, that are used to protect most modern mobile applications, are faced with usability and security problems related to their static and one-shot nature. Indeed, one-shot authentication mechanisms challenge the user at the beginning of a session leaving them vulnerable to attacks on lost/stolen devices or session hijacking. In addition, static authentication mechanisms always use the same challenges to authenticate the user without considering the dynamic nature of the risk related to the authentication context. To mitigate these challenges, we propose RLAuth, a risk-based authentication system that can automatically adapt the level of challenge presented to the user on each authentication request based on the current context. RLAuth is based on binary anomaly detection, which is solved using a deep reinforcement learning agent that acts as the classifier. To cope with the high class imbalance in the anomaly detection problem, we propose to use a balanced sampling technique during experience replay and an imbalanced correction factor during reward computation. We evaluate RLAuth on a public dataset using the G-mean metric which is the square root of the product of sensitivity with specificity. This metric is efficient to measure the classification performance of a model under class imbalance since it does not overfit to the majority class. Finally, RLAuth obtained a G-Mean of 92.62%. In addition, the reinforcement learning agent can be trained offline for acceptable results in about 130 s and can then be periodically retrained to improve its performance over time.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it