Presentation is loading. Please wait.

Presentation is loading. Please wait.

Detecting DDoS Attacks on ISP Networks Ashwin Bharambe Carnegie Mellon University Joint work with: Aditya Akella, Mike Reiter and Srinivasan Seshan.

Similar presentations


Presentation on theme: "Detecting DDoS Attacks on ISP Networks Ashwin Bharambe Carnegie Mellon University Joint work with: Aditya Akella, Mike Reiter and Srinivasan Seshan."— Presentation transcript:

1 Detecting DDoS Attacks on ISP Networks Ashwin Bharambe Carnegie Mellon University Joint work with: Aditya Akella, Mike Reiter and Srinivasan Seshan

2 ISP Perspective of DDoS Attack Victim Attacker (Incarnation II) My ISP ISP3 ISP2 Hot potato routing

3 Problem Statement How can an ISP find out if: How can an ISP find out if: Its Backbone is carrying “useless” attack traffic? Its Backbone is carrying “useless” attack traffic? Its Backbone is itself under attack? Its Backbone is itself under attack? Focus of this talk: Focus of this talk: Sketch a solution approach Sketch a solution approach Discuss the main challenges Discuss the main challenges

4 Approach Record “normal” traffic at routers; identify anomalies Exchange suspicions among routers to reinforce anomaly detection Traffic Profile Destination: 10.1.203.210 #Flows: … #Bytes: … MB Traffic Profile Destination: 10.1.203.210 #Flows: … #Bytes: … MB

5 Basic Approach 1. Record “normal” traffic at routers 2. Detect “abnormalities” in traffic Challenges a. What is normal and what is abnormal? b. Is it robust? c. How quickly can we identify deviations? d. Can it really be implemented on a backbone router? e. Response strategy?

6 Proposed Solution Maintain Traffic Profiles Each router constructs profiles of traffic Each router constructs profiles of traffic Longer time-windows  normal traffic Longer time-windows  normal traffic Smaller time-windows  current traffic Smaller time-windows  current traffic Become suspicious if current profile violates normal profile Become suspicious if current profile violates normal profile

7 Important Challenges 1. Day-of-week and Time-of-day effects Maintain per-day per-daytime statistics Maintain per-day per-daytime statistics 2. Flash crowds Example of “harmless” but infrequent event Example of “harmless” but infrequent event Attack-volume alone is not a sufficient indicator Attack-volume alone is not a sufficient indicator “Fingerprint” the destination-bound traffic “Fingerprint” the destination-bound traffic Number of sources, source-subnets, flows, distribution of flow lengths, etc. Number of sources, source-subnets, flows, distribution of flow lengths, etc.

8 Traffic Fingerprints Some examples Total traffic to destination Total traffic to destination Source subnet characterization Source subnet characterization Total number of “flows” to a destination Total number of “flows” to a destination How many /24 subnets are observed in the traffic to this destination How many /24 subnets are observed in the traffic to this destination Flow-length distribution Flow-length distribution E.g., are there a lot of small flows? E.g., are there a lot of small flows?

9 Stream Sampling Memory/computation constraints at routers Memory/computation constraints at routers Keep statistics about every destination? Keep statistics about every destination? Only for popular ones  traffic to whom exceeds a fraction  of link capacity Only for popular ones  traffic to whom exceeds a fraction  of link capacity Use sample-and-hold or multistage filters [Estan01] Use sample-and-hold or multistage filters [Estan01] Count unique subnets in a packet stream Count unique subnets in a packet stream Memory =  (size of stream)! Memory =  (size of stream)! Use F 0 computation algorithms [Alon96, Gibbons01] Use F 0 computation algorithms [Alon96, Gibbons01] Do it in much smaller (constant!!) space and time Do it in much smaller (constant!!) space and time

10 Proposed Solution Increasing Robustness Single router has only local view  can make mistakes Single router has only local view  can make mistakes Traffic perturbations due to traffic engineering Traffic perturbations due to traffic engineering False alarms! False alarms! Suppose attacker “mimics” normal traffic at a router Suppose attacker “mimics” normal traffic at a router Attack goes undetected! Attack goes undetected! Mimicking at more than a few routers within an ISP would be hard! Mimicking at more than a few routers within an ISP would be hard! Use router consensus for reinforcing suspicions across routers Use router consensus for reinforcing suspicions across routers

11 Preliminary Results Single Router Detection Accuracy Experimental Setup Abilene-II traffic trace (70 minutes) Abilene-II traffic trace (70 minutes) Samples taken across a window of about 1 minute Samples taken across a window of about 1 minute Synthetic attack traffic (trinoo, TFN, TFN2k, etc.) Synthetic attack traffic (trinoo, TFN, TFN2k, etc.) Attack Detection Accuracy False positive rates ≤ 6%, lower for “ unpopular ” destinations False positive rates ≤ 6%, lower for “ unpopular ” destinations False negative rates decrease rapidly as the “ rate ” of attack traffic increases False negative rates decrease rapidly as the “ rate ” of attack traffic increases

12 Conclusions and Future Work Conclusions Conclusions Fingerprinting traffic allows for detection of subtle attack patterns not apparent from volume alone Fingerprinting traffic allows for detection of subtle attack patterns not apparent from volume alone Distributed detection makes it harder for an attacker to mimic traffic at multiple routers Distributed detection makes it harder for an attacker to mimic traffic at multiple routers Directions for future work Directions for future work Identify various attack scenarios Identify various attack scenarios Optimize computation/space requirements Optimize computation/space requirements Consensus algorithm; convergence and effectiveness Consensus algorithm; convergence and effectiveness Validate over real attack datasets Validate over real attack datasets

13 Backup Slide Overheads Use  = 0.1  memory ~ 3600 bytes per destination Use  = 0.1  memory ~ 3600 bytes per destination Approximate number of popular destinations = 1/  Approximate number of popular destinations = 1/  where  is the fraction of link capacity where  is the fraction of link capacity 360 KB per statistic – if we use  = 1% 360 KB per statistic – if we use  = 1% Can a high-end router have a few MBs of SRAM? Can a high-end router have a few MBs of SRAM? AlgorithmsAMS96GT01 Accuracy 1+ ,  > 1 1 ± ,  > 0 Memory (bytes) 4 36/  2 Byte operations ~4 ~4~6 Counting unique items in a stream (zeroeth moment F 0 )


Download ppt "Detecting DDoS Attacks on ISP Networks Ashwin Bharambe Carnegie Mellon University Joint work with: Aditya Akella, Mike Reiter and Srinivasan Seshan."

Similar presentations


Ads by Google