In Figure 7 of the paper, we show an example of detection when neither edge/source/destination is individually anomalous but as a whole, it is a microcluster anomaly. It can similarly be detected when there are multiple web-hosting servers.
We have extended the work for multi-aspect data i.e. records with multiple features. Currently our approach is capable of handling structured data only. It should be interesting to see how to detect anomalies using this approach in data which does not have a regular structure.
We handle spatial locality in terms of not just the source but also the destination, therefore we should be able to handle DDoS like attacks when simultaneous edges come from several sources trying to deny one particular destination.
Thank you :)
1. We use a temporal decay (alpha).
2. Good question! We consider similar edges as those having at least one of source and destination node as the same.
3. Very interesting direction for future work! We can try using a variable decay instead of a fixed one to tackle the adversary.
Thanks! There is a subtle difference. We define microclusters within the category of anomalous edge detection as 'suddenly arriving groups of suspiciously similar edges' e.g. denial of service attacks in network traffic data and lockstep behavior.
Subgraph detection on the other hand is detecting anomalous subgraphs and may also not be able to detect microclusters (suddenly arriving groups of suspiciously similar edges.)
Anomaly detection in graphs is a critical problem for finding suspicious behavior in innumerable systems, such as intrusion detection, fake ratings, and financial fraud. But most of the systems in place focus either on static graphs or on entire graph snapshots if they consider dynamic (time-evolving) graphs.
However, to minimize the effect of malicious activities and start recovery as soon as possible, we need to detect anomalies in real-time or near real-time i.e. to identify whether an incoming edge is anomalous or not, as soon as we receive it. In addition, since the number of vertices can increase as we process the stream of edges, we need an algorithm which uses constant memory in graph size. Moreover, fraudulent or anomalous events in many applications occur in microclusters or suddenly arriving groups of suspiciously similar edges e.g. denial of service attacks in network traffic data and lockstep behavior.
In this work, we propose MIDAS, which detects microcluster anomalies, or suddenly arriving groups of suspiciously similar edges, in edge streams, using constant time and memory. In addition, by using a principled hypothesis testing framework, MIDAS provides theoretical bounds on the false positive probability, which earlier methods do not provide. Also, we are up to 644 times faster and 48% more accurate than previous state of the art approaches.
Given a stream of graph edges from a dynamic graph, how can we assign anomaly scores to edges in an online manner, for the purpose of detecting unusual behavior, using constant time and memory? Existing approaches aim to detect individually surprising edges. We present MIDAS, which focuses on detecting microcluster anomalies, or suddenly arriving groups of suspiciously similar edges, such as lockstep behavior, including denial of service attacks in network traffic data.
MIDAS has the following properties:
(a) it detects microcluster anomalies while providing theoretical guarantees about its false positive probability;
(b) it is online, thus processing each edge in constant time and constant memory, and also processes the data 108 − 505 times faster than state-of-the-art approaches;
(c) it provides 46%-52% higher accuracy (in terms of AUC) than state-of-the-art approaches.
MIDAS detects microcluster anomalies from an edge stream in constant time and memory, while providing theoretical guarantees about its false positive probability.
Microcluster anomalies are suddenly arriving groups of suspiciously similar edges, such as lockstep behavior and denial of service attacks in network traffic data.
Few use cases include detecting intrusions, fake ratings and financial fraud.