TY - GEN
T1 - Constrained Concealment Attacks against Reconstruction-based Anomaly Detectors in Industrial Control Systems
AU - Erba, Alessandro
AU - Taormina, Riccardo
AU - Galelli, Stefano
AU - Pogliani, Marcello
AU - Carminati, Michele
AU - Zanero, Stefano
AU - Tippenhauer, Nils Ole
N1 - Accepted Author Manuscript
PY - 2020
Y1 - 2020
N2 - Recently, reconstruction-based anomaly detection was proposed as an effective technique to detect attacks in dynamic industrial control networks. Unlike classical network anomaly detectors that observe the network traffic, reconstruction-based detectors operate on the measured sensor data, leveraging physical process models learned a priori. In this work, we investigate different approaches to evade prior-work reconstruction-based anomaly detectors by manipulating sensor data so that the attack is concealed. We find that replay attacks (commonly assumed to be very strong) show bad performance (i.e., increasing the number of alarms) if the attacker is constrained to manipulate less than 95% of all features in the system, as hidden correlations between the features are not replicated well. To address this, we propose two novel attacks that manipulate a subset of the sensor readings, leveraging learned physical constraints of the system. Our attacks feature two different attacker models: A white box attacker, which uses an optimization approach with a detection oracle, and a black box attacker, which uses an autoencoder to translate anomalous data into normal data. We evaluate our implementation on two different datasets from the water distribution domain, showing that the detector's Recall drops from 0.68 to 0.12 by manipulating 4 sensors out of 82 in WADI dataset. In addition, we show that our black box attacks are transferable to different detectors: They work against autoencoder-, LSTM-, and CNN-based detectors. Finally, we implement and demonstrate our attacks on a real industrial testbed to demonstrate their feasibility in real-time.
AB - Recently, reconstruction-based anomaly detection was proposed as an effective technique to detect attacks in dynamic industrial control networks. Unlike classical network anomaly detectors that observe the network traffic, reconstruction-based detectors operate on the measured sensor data, leveraging physical process models learned a priori. In this work, we investigate different approaches to evade prior-work reconstruction-based anomaly detectors by manipulating sensor data so that the attack is concealed. We find that replay attacks (commonly assumed to be very strong) show bad performance (i.e., increasing the number of alarms) if the attacker is constrained to manipulate less than 95% of all features in the system, as hidden correlations between the features are not replicated well. To address this, we propose two novel attacks that manipulate a subset of the sensor readings, leveraging learned physical constraints of the system. Our attacks feature two different attacker models: A white box attacker, which uses an optimization approach with a detection oracle, and a black box attacker, which uses an autoencoder to translate anomalous data into normal data. We evaluate our implementation on two different datasets from the water distribution domain, showing that the detector's Recall drops from 0.68 to 0.12 by manipulating 4 sensors out of 82 in WADI dataset. In addition, we show that our black box attacks are transferable to different detectors: They work against autoencoder-, LSTM-, and CNN-based detectors. Finally, we implement and demonstrate our attacks on a real industrial testbed to demonstrate their feasibility in real-time.
KW - Adversarial Machine Learning
KW - Autoencoder
KW - Classifier Evasion
KW - Deep Learning
KW - Evasion Attack
KW - Industrial Control System
KW - Intrusion Detection
KW - Mean Squared Error
KW - Multivariate Time Series
UR - http://www.scopus.com/inward/record.url?scp=85098066115&partnerID=8YFLogxK
U2 - 10.1145/3427228.3427660
DO - 10.1145/3427228.3427660
M3 - Conference contribution
AN - SCOPUS:85098066115
T3 - ACM International Conference Proceeding Series
SP - 480
EP - 495
BT - Proceedings - 36th Annual Computer Security Applications Conference, ACSAC 2020
PB - ACM
T2 - 36th Annual Computer Security Applications Conference, ACSAC 2020
Y2 - 7 December 2020 through 11 December 2020
ER -