Network Security Lab (NSL)
We are primarily interested in developing a concrete mathematical framework for bug-free algorithm design in security as well as in studying security protocols from communication, coding and information theoretic view points. We utilize tools from these well founded areas to design and analyze security protocols as well as identify bugs that may go undetected otherwise.
Our Contributions in Emerging Wireless Networks:
- Modeling Vulnerabilities: In order to protect user information and provide robust network operation, a network designer must know what types of attacks to protect against and what threats the network must be robust against. Hence, a necessary prerequisite to robust protocol design is to model network vulnerabilities and the impact of malicious attacks on the network. Protecting the information exchange between users is a problem that is well studied in cryptography, but the introduction of an entire network of users into the scenario introduces additional vulnerabilities when the information security is jointly considered with network protocols such as routing and multiple access. Furthermore, the propagation of information throughout a large network allows a locally-constrained adversary to have a more significant impact on information and network security, as the effects of the adversary's attack can quickly spread beyond the local neighborhood. We have investigated vulnerabilities in large-scale networks due to node capture attacks, in which an adversary physically compromises nodes in the network to extract secret information and gain network presence, and jamming attacks, in which an adversary intentionally interferes with wireless communication in the network to prevent information access and transport.
- Secure Localization: Wireless sensor networks, which represent a basic tenet of what we call ubiquitous computing, are now or will soon be deployed in physical environments that are vulnerable not only to the vicissitudes of nature but also to acts that could be easily viewed as hostile attacks by potent adversaries. Indeed, unattended operation of sensor-network nodes in hostile environments requires that we rethink the definition of our adversary, its capabilities and modes of attack. There are few problems of wireless sensor network design and analysis that areas challenging as localization and time synchronization. Yet both are fundamental building blocks not just for new applications and but also security services themselves. The natural interplay between space and time measurements and bounds, which are basic to both localization and time synchronization, produces a largely uncharted research territory. And, of course, the new capabilities and attack modes of the new adversary complicates the landscape in unanticipated ways. We developed robust location estimation approach called SeRLoc for wireless sensor networks that need to operate in hostile environments. We also developed a high resolution version of the localization. This work has eventually led to technology transition to Navy. We also organized a workshop and edited a book in this area. Applications of secure location estimation include relative location estimation, building and infrastructure health monitoring, location verification in cognitive radios.
- Network Performance Analysis: One of the primary tasks of wireless sensor networks is to monitor a Field of Interest (FoI). The availability of observations is directly related to the number of sensors able to sense a particular event, and can be quantified by computing the fraction of the FoI covered by at least a threshold number of sensors, also know as k-coverage. Previous work on evaluating the k-coverage, assumed that sensors have identical sensing areas and/or conform to the idealized unit disk model. However, we consider sensors of multiple sensing modalities such as acoustic, optical, infrared, CCD, magnetic, or thermal, have sensing areas significantly different than the unit disk model and may be concurrently deployed, thus forming a heterogeneous WSN. Alternatively, for applications such as area surveillance and habitat monitoring the network performance is related on how well the deployed network can monitor mobile targets that cross the FoI. We quantify the latter by computing the probability of detecting a target crossing the FoI. As in the case of k-coverage, analytically computing the target detection probability assuming a heterogeneous WSN is a challenging problem.
- Energy-Aware Secure Group Communications in Wireless/Sensor networks: We address the problem of securing multicast communications in an energy-constrained ad-hoc network environment. Existing efficient key distribution schemes for wired networks that rely on logical hierarchies are extremely energy inefficient for energy-constrained wireless ad-hoc networks. The joint consideration of routing and physical layer algorithms is critical for developing energy-efficient key distribution schemes. By formulating the problem we showed that solution is hard to compute. Therefore we developed greedy, routing-aware key-distribution algorithms that are easy to compute.
- Vehicular Network Security: The Vehicular Ad hoc Network (VANET) is an emerging wireless network that provides easy access to the vehicular communications, allowing eavesdroppers to estimate vehicle locations based on communication signal properties, such as signal strength. Therefore, the successful deployment of VANET infrastructure, such as the US Department of Transportation Vehicle-Infrastructure-Integration project, cannot be realized without addressing the security and privacy challenges. We showed that the requirements of VANET applications and the restricted mobility properties allow vehicles to be continually tracked, and that the location tracking of a target vehicle can lead to the compromise of privacy of the user in that vehicle. Based on observed properties of mobile nodes in VANET, such as the group navigation of vehicles and the uncertainty of vehicle movement when merging/changing lanes, we proposed defense mechanisms that can protect privacy of vehicle users. Our preliminary work was presented at the workshop on Embedded Security in Cars (ESCAR) 2005.
Our Contributions in Industrial and Societal Applications:
- Medical Security: Privacy of patient health records has become a pressing societal issue, leading to the Health Insurance Portability and Accountability Act (HIPAA) in April 2003. For electronic health records (EHR), privacy is ensured by restricting the access to authorized users only at any time. Following the HIPAA, efforts have been devoted to studying problems of controlling the access to EHR and ensuring confidentiality during the retrieval from a database. We identified that there was a patient privacy vulnerability that was not addressed by HIPAA and by any of the prior work in EHR. We showed that the authorized users can illegally distribute EHR to unauthorized parties, after reception of the EHR. We then developed methods for enhancing post-reception privacy of EHR by enabling tracing medical images in a multi-user communication environment. Using our proposed fingerprinting techniques, the source of an illegal EHR leakage can be identified in a resource-efficient way. Our scheme is highly robust to typical medical image processing and collusion attacks, while yielding high quality watermarked images. Our work has appeared in Diagnostic Imaging Magazine's PACSweb as a news article published on October 25, 2005, titled "Image watermarking could address HIPAA loophole" .
- Airplane Security: Future advances in aviation industry will include the e-enabled airplanes that possess wired and wireless networking capabilities allowing for the electronic distribution of airplane loadable software and data, e.g. airplane health data, leading to significant time and cost benefits. However, as e-enabled airplanes connect with network environments off-board, vulnerabilities in open networks and use of commercial-off-the-shelf IT components present opportunities for security attacks. Some airplane loadable software, e.g. flight control computer software, have safety implications and the integrity of such safety-critical parts must be protected at all times. On the other hand, the integrity of the non-safety-critical airplane software also needs to be protected to ensure passenger comfort and confidence on airline business processes, and to reduce unwarranted flight delays. Based on a generic heterogeneous system for electronic storage and distribution of airplane software, we identified security threats to airplane safety and operation and maintenance and proposed security primitives to address them. This research is conducted in collaboration with Boeing Phantom Works and is a Boeing initiative towards becoming part of standardization efforts in the aviation industry. We have presented our work as a position paper at the NSF sponsored National Workshop on Aviation Software Systems in October 2006.
- Electronic Voting Security: Electronic voting is a pivotal social application of cryptographic protocols. However, the use of insecure Internet, well documented cases of incorrect implementations, and the resulting security breaches reported recently have significantly lowered public confidence in the security, reliability and privacy of e-voting. Although e-voting schemes have been developed over the last two and a half decades, their requirements are seen to be conflicting and there is no existing framework under which they can be studied and compared. In order to streamline the ongoing effort to improve public confidence in e-voting, we proposed such a framework that shows which requirements can be simultaneously satisfied by different classes of e-voting schemes, allowing designers to check the conditions their proposed schemes satisfy and the tradeoffs, thus reducing unwarranted claims or unintended errors. We have illustrated the use of our framework by analyzing some of the existing e-voting schemes. Furthermore, we have also proposed a framework for understanding Mixnets that are designed to use multi-stage cryptography and permutation to protect communicated electronic voting results as well as other types of user information such as medical records and personal opinions.
Our Contributions in Internet:
- Internet Security Standards: Cipher Block Chaining MAC (CBC-MAC) that was recommended by the National Institute for Standards and Technologies (NIST) for block cipher modes of operation is known to have security weaknesses. Therefore, in order to overcome the known security weaknesses of CBC-MAC under variable input message sizes, we proposed the AES-CMAC based on Cipher-based Message Authentication Code (CMAC) and with Advanced Encryption Standard (AES) algorithm, an approved block cipher in the Federal Information Processing Standard Pub. 197., as the underlying block cipher. We employed AES-128 with block size of 128 bits, and proposed the sub-key generation, MAC generation and verification algorithms for AES-MAC. Additionally, based on AES-MAC, we propose AES-MAC-96 and AES-MAC-PRF-128 for IP Security (IPsec) implementations. Our work on AES-CMAC algorithm has been published under the standards track of the IETF as RFC 4493, RFC 4494. Further, we have also specified the AES-MAC-PRF-128 for pseudo-random-functions with fixed as well as variable key sizes in RFC 4615.
- Efficient Key Distribution for Secure Multicast Communications for Internet: Internet is a suitable global network for digital video broadcast, distributed gaming and multimedia data streaming. We have studied point-to-multipoint or the single sender- multiple receiver model of secure multicast and developed analytical tools for design and analysis of the secure multicast. In particular, we showed that tree-based key distributions can be analyzed using well founded results from Information Theory and showed that the average number of cryptographic keys to be stored by a user in such a system is characterized by the entropy of member key update process under member deletion (which also includes the voluntary de-registration of a member). Current research involves developing explicit design methods for these schemes. As part of our research, we showed that a "secure" multicast protocol by IBM satisfied optimal key allocation principle but led to user collusion.
- Secure Multimedia Multicast: For providing secure multimedia multicast for handheld devices, we developed a key distribution method based on residues that allows reuse of a member's secret in key distribution. The scheme has been incorporated into data embedding for key updates. The interest now is in developing techniques that will combine multimedia features and scalable key distribution techniques to develop more efficient key distribution schemes. Research directions also include securing speech and music in limited scenarios.
- Secret Sharing Schemes: A (t;n) threshold scheme allows a secret to be distributed among a group of n participants in such a way that t or more participants can construct the secret by pooling their shares, but the secret remains undetermined to (t - 1) or fewer participants. Threshold schemes find vast application in collective control, threshold cryptography, distributed secret key storage, file sharing etc. Our research focuses on how to maintain the threshold via public broadcast channel when disenrolling untrustworthy participants, i.e., threshold schemes with disenrollment capability (first defined by Blakley et al). We showed that the original definition has a potential problem of rendering the dealer no control over disenrollment, and resolved the problem by proposing a more secure model of threshold scheme with disenrollment. We established a lower bound on the size of broadcast messages in the new model and study the tradeoff between share size and broadcast message size.
Our current research work is generously supported by the following:
Professor Poovendran's Ph.D. work was supported by Information Assurance Group of NSA as well as the Computer and Communication Division of the US Army Research Laboratory.
Professor Poovendran is looking for graduate students with one or more of the following backgrounds: (a) signal and image processing course work at the graduate level with project experience with image coding (b) networking, (c) computer hardware with experience in dsp programming, (d) applied mathematics, (e) statistics. If you are a prospective student with background in network security and cryptography, that will be a plus point. For questions related to admission to graduate program contact the EE Admissions. Some useful information is also available at scholarship resources for UW students. You may want to check that list for potential sponsors. He also works with undergraduate students who are motivated. If you are an undergraduate student who is interested in working with him, please meet with an EE undergraduate advising officer before approaching him for mentoring or research supervision.