Three theories for delays, clocks and security in wireless networks
P. R. Kumar · 2012
We propose three theories, which can be regarded as attempts to characterize and establish guaranteed properties of wireless networks: (i) How and to what extent can we deliver packets with hard delay bounds? (ii) How and to what extent can we synchronize clocks in wireless networks? (iii) Can we develop provably secure protocols for the entire life-cycle of wireless networks that also optimize a utility measure while operating in a hostile environment? For the first problem, consider an access point serving several clients over unreliable wireless links. Suppose packets arrive for/from the clients, with each packet having a hard deadline, after which it is dropped. We characterize precisely the mix of delivery ratios, channel unreliabilities and hard deadline that the access point can guarantee, under some models. For the second problem, consider a wireless network where clocks at the nodes are linear, though with different rates (skews) and offsets. Nodes can exchange packets with their neighbors, with direction dependent delays. We characterize precisely to what extent clocks can and cannot be synchronized and delays determined. Under a random model the end-to-end error can be kept bounded irrespective of network size. Concerning the third problem, traditionally, wireless protocols have been developed to provide performance. As attacks are identified, the protocols are fortified against the identified vulnerabilities. However, holistic guarantees are not provided against other attacks. We seek to reverse this paradigm. We propose a provable approach that guarantees the protocol suite is secure when the nodes are subject to certain assumptions. The protocols take a set of good nodes mingled with unknown malicious nodes from primordial birth to an operating network, while attaining min-max of a utility function. The maximization is over protocols announced and followed by the good nodes, and the minimization is over all behaviors of the malicious nodes. Further, the malicious nodes are reduced to either cooperating or jamming. [Joint work with Vivek Borkar, Nikolaos Freris, Scott Graham, I-Hong Hou, Yih-Chun Hu, Jonathan Ponniah and Roberto Solis].