haku: @keyword anomaly detection / yhteensä: 10
viite: 5 / 10
Tekijä: | Alene, Henok |
Työn nimi: | Graph based clustering for anomaly detection in IP networks |
Julkaisutyyppi: | Diplomityö |
Julkaisuvuosi: | 2011 |
Sivut: | [11] + 67 Kieli: eng |
Koulu/Laitos/Osasto: | Tietotekniikan laitos |
Oppiaine: | Informaatiotekniikka (T-61) |
Valvoja: | Oja, Erkki |
Ohjaaja: | Hätönen, Kimmo |
OEVS: | Sähköinen arkistokappale on luettavissa Aalto Thesis Databasen kautta.
Ohje Digitaalisten opinnäytteiden lukeminen Aalto-yliopiston Harald Herlin -oppimiskeskuksen suljetussa verkossaOppimiskeskuksen suljetussa verkossa voi lukea sellaisia digitaalisia ja digitoituja opinnäytteitä, joille ei ole saatu julkaisulupaa avoimessa verkossa. Oppimiskeskuksen yhteystiedot ja aukioloajat: https://learningcentre.aalto.fi/fi/harald-herlin-oppimiskeskus/ Opinnäytteitä voi lukea Oppimiskeskuksen asiakaskoneilla, joita löytyy kaikista kerroksista.
Kirjautuminen asiakaskoneille
Opinnäytteen avaaminen
Opinnäytteen lukeminen
Opinnäytteen tulostus
|
Sijainti: | P1 Ark Aalto 7071 | Arkisto |
Avainsanat: | anomaly detection graphs node degree node clustering weight matrix partition clustering |
Tiivistelmä (eng): | In IP networks, an anomaly detection system identifies attacks, device failures or other unknown processes that deviate from the normal behaviour of the network known as anomalies. The thesis studied anomaly detection in traffic datasets from IP networks. The datasets contained high number of normal events and few anomalies. This resembles a normally operating network. We construct graphs from traffic data and study their properties. We formulated anomaly detection as a graph based clustering problem. A novel graph bi-partitioning algorithm called NodeClustering was designed to separate normal samples from anomalous ones. Performance of NodeClustering was investigated with extensive network traffic data. The performance was compared with state of the art graph based spectral clustering algorithms. NodeClustering identified all the known intrusions in the data and outperformed the compared graph based methods with an average improvement of 50% on the true positive rate with lowest false positive rate on the studied datasets. In addition, its applicability to one non-traffic dataset was shown. NodeClustering can be used in IP networks to detect anomalies. In the future, threshold used for graph partitioning can be studied further and computationally efficient < methods to construct larger graphs might be studied. |
ED: | 2011-12-14 |
INSSI tietueen numero: 43255
+ lisää koriin
INSSI