Anomaly Based Intrusion Detection in Distributed Applications without global clock Eric Totel, Mouna Hkimi, Michel Hurfin, Mourad Leslous, Yvan Labiche SEC2-2016 5 July 2016
Outline of the Presentation • Position of the problem • Building a distributed application behavior model • Partial Event Ordering • Automaton recognizing sequences • Temporal properties • Applying Detection on an example • Results on a distributed file system application CentraleSupelec 2
Intrusion Detection in Distributed Systems • Several nodes running processes • Intrusion Detection Systems are deployed • On the network (NIDS) • On each node (HIDS) • Local Detection of compromission • No relationship between the states of the several nodes • Alerts emitted takes into account the state of one node • Current solutions • Alert correlation: Requires total ordering of all alerts • DIDS: Requires total ordering of all events analyzed • In Cloud environments: virtual machines are often desynchronized (clock drift) CentraleSupelec 3
The Case of a Distributed Application • How to enhance the detection ? • Statement • The states of the different nodes are not independent • As such the behaviors of the different nodes of the application are not independent • The actions performed by the nodes are causally dependent on each other • Local actions • Messages exchanged • Solution • Build a reference model that takes into account the causal dependencies between the nodes • Without relying on a total ordering of the events (no global clock) CentraleSupelec 4
Logs and partial ordering • On each node, a process produces a total ordered log • Partial Ordering of the events on different nodes (Lamport happened before relationship) 8 e 2 E α , 8 f 2 E α , e � α f • e occured before f in the same log E α i • e is a message send and f its receipt e � α g and g � α f • there exists g such that • How to learn the right sequences of actions performed by the distributed processes ? CentraleSupelec 5
Example of Logs • A trace: two logs of two processes and a b c1!m Execution α p1 E α E α 1 2 1 a d p2 2 b c1?m d c1?m e 3 c1!m e • On this execution, a � α b • No order relation between b and d: • {a, b, d, c1!m, c1?m, e} is a valid sequence … but not the only one ! CentraleSupelec 6
Notion of a valid sequence • Observed correct normal sequence • Compliant with the partial relationship • A sequence of events is valid iff CentraleSupelec 7
Generation of valid sequences (1) p2 Execution α Generating the lattice of e consistent cuts E α E α 1 2 1 a d c1?m 2 b c1?m A valid sequence is a sequence 3 c1!m e of events consumed by a path d in the lattice of consistent cut a b c1!m p1 CentraleSupelec 8
Generation of valid sequences (2) Generation of an automaton containing all the paths in the lattice of consistent cuts p2 c1!m 3 d e 2 b c1?m e d 13 9 6 c1!m 1 a d 7 0 d c1?m b 17 a 16 d a b c1!m p1 CentraleSupelec 9
Automaton from several executions Execution α Execution β E β E β E α E α 1 2 1 2 1 a d a f 2 b c1?m c1!m c1?m 3 c1!m e g � ���� � � � ���� � � � � � ���� � � � � � � � � � � � ���� �� � �� ���� � � � �� �� �� ���� Merge the start states �� of all the automata � �� CentraleSupelec 10
Analysis of the automaton • Contains only the observed valid sequences • In practice: • In a heavy distributed application, it is very difficult to exhibit all the behaviors of the application due to concurrency • It is thus very difficult to learn a complete behavior model • Solution: • Generalization of the automaton • Permits to introduce new unlearned behaviors • Ensures that all the original valid sequences are included in the generalized automaton CentraleSupelec 11
Generalization (k-tail algorithm) ���� � � � � ���� � � � � � ���� � � � � � � � � Disadvantage: can introduce incorrect sequences � � � ���� �� of events at the same time � �� ���� � � � �� �� �� ���� �� � k=1 (a low k permits a higher generalization) �� ���� � � � � � � � � ���� � � Advantage: can introduce new valid unlearned � � � ���� � � � � ���� � � sequences of events � �� � � � � �� ���� �� � CentraleSupelec 12
How to deal with incorrect sequences ? • Duality of models • Automaton: exhaustive list of sequences • Temporal properties: properties on the types of events • Temporal invariants • Issued from the domain of test • Three invariants considered (a and b are event types) • a is always followed by b • a is never followed by b • a always precedes b CentraleSupelec 13
Invariants on our example ���� � � � � ���� � � � � � ���� � � � � � � � � � � � ���� �� � �� ���� � � � �� �� �� ���� �� � �� (total of 59 invariants) Generalization Model checking ���� � � � � � � � � ���� � � � � � ���� � � � ���� � � � � �� � � � � �� ���� �� � CentraleSupelec 14
Duality of models Model Generalized Invariants that can be Automaton violated by the generalized automaton ���� � � � � � � � � ���� � � � � � ���� � � � � ���� � � � �� (total of 10 invariants) � � � � �� ���� �� � Non acceptable sequence {a, b, c1!m, d, c1?m, g} CentraleSupelec 15
Valid/Accepted/Acceptable sequences • Invariants computed on the original lattice of consistent cuts ∑ '' acceptable • Invariants on valid sequences of sequences events • Invariants are less restrictive than the automaton ∑ ' • We consider a sequence is sequences acceptable if it is accepted by the accepted by the generalized ∑ automaton and complies with the automaton valid invariants sequences CentraleSupelec 16
Detection algorithm • Given a trace • Is this trace compliant: • With the generalized automaton • With the temporal invariants • Two strategies • All total ordering of the events of the trace are compliant with the model • At least one order of the events of the trace is compliant with the model • In practice • Strategy « all » is more time consuming • Similar false positive rate in both approaches CentraleSupelec 17
Simple Example: e-commerce • 3 processes: article buying, 70 possible different behaviours P2-P1!SEARCH P1-P2?AVAILABLE P1-P2?AVAILABLE P2-P1!BUY P1-P2?SOLD Process (p2) P1-P3!SOLD P3-P1?BUY P3-P1?SEARCH P1-P3!AVAILABLE Server (p1) P2-P1?BUY P1-P2!SOLD P2-P1?SEARCH P1-P2!AVAILABLE Process (p3) P3-P1!SEARCH P1-P3?AVAILABLE P3-P1!BUY P1-P3?SOLD CentraleSupelec 18
Detection Accuracy • Simulations of an intrusion • Removing an event • Modifying the order of events • Adding new events • Violating the integrity of the distributed logs • Are detected by the approach CentraleSupelec 19
Generalization and False Positive Rate False Positive Rate • Learning Phase with 10, 20, 30, 40, 50, 60 traces traces learned=10 traces learned=20 traces learned=30 traces learned=40 traces learned=50 traces learned=60 • With a generalization 90% 85% 84% 80% parameter k=1, 2, 3, 4, 5 75% 71% 70% 70% 68% 65% 60% 52% 50% • Result: 42% 40% 39% • The generalization 31% 30% 30% decreases the rate of the 22% 20% 19% 19% 18% 16% false positives, even with a 10% 10% 9% 9% 8% 8% 6% low number of traces learnt 3% 3% 2% 1% 0% 0% 0% 0% 1 2 3 4 5 k CentraleSupelec 20
Real World Evaluation: XtreemFS • High Availability Distributed Replicated File System • Intrusion Detection approach applied on a simple configuration of the nodes CentraleSupelec 21
Experimentation applied • Writing of a set of files • 500 files used to learn the model • 1640 files written to measure the false positive rate • Traces obtained on each node by instrumenting the code of the file servers • One trace for a complete file write CentraleSupelec 22
Model Size • Number of traces used Model Size to learn the model 7800 800 grow 7600 700 • The number of Number of invariants 7400 600 Number of States invariants lower 7200 500 • The size (number of invariants 7000 400 states) of the 6800 300 States automaton grows (k- 6600 200 tail applied with k=1) 6400 100 6200 0 10 50 100 200 300 400 500 Number of Traces CentraleSupelec 23
Recommend
More recommend
