Collective anomalies co-occurrence of data instances in a particular form that makes them anomalies
co-occurrence of data instances
in a particular form that makes them anomalies
Relationship between data instances
Detecting anomalous sequence
Sequences might not be of equal length
Sequences may not be aligned with each other
Are sequences aligned ?
Transform the sequences to a finite feature space and then use a point anomaly detection technique
in the new space to detect anomalies
time-based inductive learning
generate rules from the set of normal sequences
Finite State Automatons (FSA)
Hidden Markov Model
Probabilistic Suffix Trees (PST)
Sparse Markov Trees (SMT)
Does normal behavior of the time-series follows a defined pattern ?
Window Comparison Anomaly
Maximum Entropy Markov Models (Maxent)
Conditional Random Fields (CRF)
length of the anomalous subsequence to be detected is not generally de¯ned
create a robust model of normalcy
Detecting anomalous subgraphs in a given large graph
measure the regularity or entropy of the sub-graph in the context of the entire graph to determine its anomaly score
analyze the frequency of a subgraph in the given graph to determine if it is an anomaly or not.
Finding subgraphs or subcomponents in the data that are anomalous.