Temporal Leakage
A phenomenon in machine learning evaluation in which, because the data split does not respect temporal order, future information inadvertently enters the training set, thereby overestimating model performance.
Timeline (last 90 days)
Intrusion detection evaluations on standard benchmarks such as CSE-CIC-IDS-2018 are often inflated due to temporal leakage caused by overlapping sliding windows and data contamination caused by fitting preprocessing before the train/test split.
The paper points out that pretrained embeddings may leak temporal information, requiring a season-based split to prevent data leakage.
The SkinCast team cuts off temporal leakage through a season-based split, that is, training on data from earlier seasons and evaluating on data from later seasons.
Data splits that ignore temporal order introduce information leakage, thereby inflating model performance.
The spatiotemporal split requires that the training and test sets do not overlap in either time or space, simultaneously avoiding temporal leakage and spatial leakage.
All Facts (5)
Intrusion detection evaluations on standard benchmarks such as CSE-CIC-IDS-2018 are often inflated due to temporal leakage caused by overlapping sliding windows and data contamination caused by fitting preprocessing before the train/test split.
50%UnverifiedThe paper points out that pretrained embeddings may leak temporal information, requiring a season-based split to prevent data leakage.
50%UnverifiedThe SkinCast team cuts off temporal leakage through a season-based split, that is, training on data from earlier seasons and evaluating on data from later seasons.
50%UnverifiedData splits that ignore temporal order introduce information leakage, thereby inflating model performance.
50%UnverifiedThe spatiotemporal split requires that the training and test sets do not overlap in either time or space, simultaneously avoiding temporal leakage and spatial leakage.
50%