Alexandra Czarlinska

Reliable Event-Detection in Wireless Visual Sensor Networks Through Scalar Collaboration and Game-Theoretic Consideration

In this work we consider an event-driven wireless visual sensor network (WVSN) comprised of untethered camera nodes and scalar sensors deployed in a hostile environment. In the event-driven paradigm, each camera node transmits a surveillance frame to the cluster-head only if an event of interest was captured in the frame, for energy and bandwidth conservation. We thus examine a simple image processing algorithm at the camera nodes based on difference frames and the chi-squared detector. We show that the test statistic of the chi-squared detector is equivalent to that of a robust (non-parametric) detector and that this simple algorithm performs well on indoor surveillance sequences and some, but not all, outdoor sequences. In outdoor sequences containing significant changes in background and lighting, this simple detector may produce a high probability of error and benefits from the inclusion of scalar sensor decisions. The scalar sensor decisions are, however, prone to attack and may exhibit errors that are arbitrarily frequent, pervasive throughout the network and difficult to predict. To achieve attack prediction and mitigation given an attacker whose actions are not known a priori, we employ game-theoretic analysis. We show that the scalar sensor error can be controlled through cluster-head checking and appropriate selection of cluster size n. Given this attack mitigation, we employ real-life sequences to determine the total probability of error when individual and combined decisions are utilized and we discuss the ensuing ramifications and performance issues.

Attacks on Sensing in Hostile Wireless Sensor-Actuator Environments

Wireless Sensor Networks (WSNs) deployed in hostile environments are susceptible to various attacks directed at their data. In this work we focus on an emerging and largely unexplored issue arising from the presence of actuator (or actor) nodes in the form of Wireless Sensor Actuator Networks (WSANs). Specifically, we consider the case where hostile WSAN nodes belonging to a foreign network directly perturb the readings of WSN nodes during sensing. The attack is modeled as affecting the decision that a WSN node reports about the presence or absence of a phenomenon to its cluster head. To assess the potential loss of sensing fidelity due to the opposing network, we employ a game theoretic analysis. We focus on determining the probability that the WSN cluster head becomes alerted to such an attack given some statistical information about the phenomenon. Our results show that an actuation attack may go unnoticed even if such an attack is not coordinated among the hostile WSAN nodes. Importantly, the number of WSN nodes in a cluster affects the probability of WSAN attack success. For clusters consisting of only a few nodes, the hostile WSAN may achieve a stealthy attack with a wider range of attack parameters. We also determine that natural phenomena with certain characteristics are more susceptible to the attack and require further sensing-verification mechanisms.

Distributed actuation attacks in wireless sensor networks: implications and countermeasures

This paper investigates the loss of sensing fidelity in a wireless sensor network resulting from a proposed novel attack. The active attack is carried out by a distributed malicious sensor actuator network (mSAN) which is able to actuate or change sensed parameters of the surrounding environment under observation. We show how the attack effectively produces a denial of service on the sensing (DoSS) of a legitimate network, causing it to observe and record false intelligence about the environment. We demonstrate how a controlled level of random mobility in the network counters the attack under various deployments, network densities and actuation radii. We conclude that a random uniform distribution may be most resilient against these attacks and that a strictly deterministic grid deployment may be most vulnerable under certain circumstances. In general we note that in physically hostile environments where sensing fidelity is important, node location becomes as sensitive for dependability purposes as encryption information

On privacy and security in distributed visual sensor networks

There is a critical need to provide privacy assurances for distributed vision-based sensor networking in applications such as building surveillance and healthcare monitoring. To effectively address protection and reliability issues, secure networking and processing must be considered from system inception. This paper presents attacks that affect the data privacy in visual sensor networks and proposes privacy-promoting security solutions based on opponent detection via game-theoretic analysis and keyless encryption.

Attack vs. failure detection in event-driven wireless visual sensor networks

In event-driven wireless Visual Sensor Networks (wVSNs), video nodes have access to additional data from scalar-sensors such as temperature or motion. The scalar-data may be used locally by the nodes instead of (or in conjunction with) vision technologies to control the potentially energy-costly transmission and storage of video frames and must thus be reliable. In this work we focus on the detection of occasional errors in such scalar-data sensors under both the scenario of harsh environmental conditions, and the scenario of hostile conditions involving an attacker. In the hostile case, the attack statistics may not be known to the cluster-head performing the error detection. We hence propose the use of a count detector in conjunction with Nash equilibrium analysis for the hostile case. We compare the detection performance of the count detector in hostile conditions to the performance of the optimal Neyman-Pearson (NP) detector which may be used under harsh conditions (scenario where the error statistics may be estimated). Through analysis and simulations we conclude that in this severe regime of attack with missing statistics, the count detector performs reasonably well compared with the optimal NP detector with significance for reliable event-driven wVSN.

G-E-M sensor networks for mission critical surveillance in hostile environments

The gathering of surveillance data such as visual intelligence from potentially hostile areas has long played a pivotal role in attaining various safety and security objectives. The methodology of gathering such surveillance is increasingly shifting towards rapid-deployment autonomous networks that limit the need for human exposure, and that cover large unattended areas while operating over extended periods of time. To achieve the surveillance objectives, such networks must be dependable and secure even in the presence of a potentially hostile counter-surveillance opponent. In this work we explicitly model and consider the presence of such an opponent in the form of a hostile sensor network with eavesdropping and actuation capabilities. We present a methodology for addressing the security and dependability issues arising in such extreme settings, which we collectively refer to as G-E-M. Specifically, we wish to ensure the legitimacy and authenticity of the gathered (G-E-M) visual surveillance in the presence of a hostile network engaged in stealthy disinformation activities. We also wish to ensure that the collected surveillance can be encrypted (G-E-M) for transmission even if keys between the nodes and the sink are temporarily compromised or otherwise unavailable. Finally we wish to ensure that the network design both inherently prolongs the lifetime of the network and also mitigates (G-E-M) deliberate energy drains. These issues are not typically examined collectively though the dependability of all these components is required to maintain the functionality and longevity of the network. Though developed and presented for the case of an attacker in the form of a hostile network, the methodologies have applicability to networks with a subset of subverted nodes that behave maliciously.

Towards characterizing the effectiveness of random mobility against actuation attacks

Actuation functionality in a sensor network enables an unprecedented interaction with the physical environment. When used by a malicious distributed network however, actuation may become a potent new attack. In this work we explore a new general class of actuation attacks which aim to disable the sensing fidelity and dependability of a wireless sensor network. We propose a countermeasure to this Denial of Service on Sensing (DoSS) based on a controlled level of random mobility. We show how the level of mobility may be traded-off to suit security needs and energy constraints, and to exploit a priori knowledge of the environment. We demonstrate how this random mobility approach performs under various strengths, densities and distributions of the two networks and show that it reduces the number of affected nodes exponentially over time. Furthermore we discuss how this simple mobility approach renders the network more fault-tolerant and resilient in an inherent way without a need for the nodes to communicate and aggregate their sensed data.

Event-Driven Visual Sensor Networks: Issues in Reliability

Event-driven visual sensor networks (VSNs) rely on a combination of camera nodes and scalar sensors to determine if a frame contains an event of interest that should be transmitted to the cluster head. The appeal of event-driven VSNs stems from the possibility of eliminating non-relevant frames at the source thus implicitly minimizing the amount of energy required for coding and transmission. The challenges of the event-driven paradigm result from the vulnerability of scalar sensors to attack or error and from the lightweight image processing available to the camera nodes due to resource constraints. In this work we focus on the reliability issues of VSNs in the case of global actuation attacks on the scalar sensors. We study the extent to which various utility functions enable an attacker to increase the average expected number of affected nodes with a relatively small penalty in the loss of stealth. We then discuss tradeoffs between different attack detection strategies in terms of the cost of processing and the required information at the cluster head and nodes.

Coordination and Selfishness in Attacks on Visual Sensor Networks

Event-driven visual sensor networks consist of collaborating camera nodes and scalar sensors which aid in the detection of events of interest in the environment. This collaboration is significant since the camera nodes generally utilize lightweight image processing in order to determine if a frame is relevant to the given application. The reliability of the supporting scalar sensor however may be compromised by an actuation attack which perturbs the sensors measurements. In this work we examine the achievable actuation of hostile nodes that are not globally coordinated and that may be selfish or untrustworthy in their preferences. We compare our findings with existing research which assumes that all the hostile nodes are coordinated to actuate with the same parameter. We determine that given certain conditions, local optimization may actually result in a stronger stealthy attack than the global coordination case.