About

Bhaskar Krishnamachari is a Professor and Ming Hsieh Faculty Fellow in Electrical Engineering at the Viterbi School of Engineering, University of Southern California. He received his B.E. in Electrical Engineering from The Cooper Union and his M.S. and Ph.D. in Electrical Engineering from Cornell University. He has been a faculty member in the Department of Electrical Engineering since 2002 and also holds a joint appointment in the Department of Computer Science. He is the Director of the Autonomous Networks Research Group and Co-Director of the Ming Hsieh Institute. His research interests are focused on the design and analysis of algorithms, protocols, and applications for next-generation wireless networks, internet of things, distributed systems, blockchain technologies, artificial intelligence, machine learning, and network economics. His work spans from theoretical analysis of algorithms to prototype software implementations of network protocols and applications. He has co-authored over 300 technical articles and has received numerous awards, including being named an IEEE Fellow in 2023, and recognition in Popular Science Magazine's 'Brilliant Ten' list in 2015 and the TR-35 list in 2011. He has also received awards such as the ASEE Terman Award, IEEE-HKN Outstanding Young Electrical and Computer Engineer Award, and the NSF CAREER award. Additionally, he has served as an editor for several prominent journals and co-edited special issues on sensor network algorithms and learning-based decision making. He authored a textbook titled 'Networking Wireless Sensors' and is co-editor of the 'Handbook on Blockchain'.

Research topics

• Computer Science
• Computer Security
• Medicine
• Data science
• Telecommunications
• Artificial Intelligence
• World Wide Web
• Simulation
• Computer engineering
• Internal medicine
• Real-time computing
• Algorithm
• Biology
• Computer network
• Virology
• Mathematics

Selected publications

• Target-Rate Least-Squares Power Allocation over Parallel Channels

  Open MIND · 2026-03-06

  preprint1st authorCorresponding

  We study power allocation over $N$ parallel Gaussian channels, such as OFDM subcarriers, when each channel has a desired target spectral efficiency. Given channel gain-to-noise coefficients $a_i>0$ and per-channel targets $T_i\ge 0$, we minimize the total squared rate deviation $\sum_{i=1}^{N}(\log_2(1+a_iP_i)-T_i)^2$ subject to a sum-power constraint $\sum_i P_i \le P_{\mathrm{tot}}$ and nonnegativity $P_i \ge 0$. We prove that the optimal allocation never overshoots any target and may leave power unused when all targets are jointly feasible, a structure fundamentally different from classical waterfilling. Using the KKT conditions, we derive a per-channel closed-form solution in terms of the Lambert~W function on the active set and reduce the remaining computation to a one-dimensional monotone bisection for the dual variable. The resulting algorithm runs in $O(N\log(1/\varepsilon))$ time and achieves up to 1{,}890$\times$ speedup over general-purpose numerical solvers at $N=1024$ channels. Numerical experiments over Rayleigh fading channels confirm that the closed-form solution matches numerical optimization to machine precision and demonstrate superior target-tracking performance compared to waterfilling, uniform allocation, and proportional fairness across a range of operating conditions.

  DOI

• From Gaussian Fading to Gilbert-Elliott: Bridging Physical and Link-Layer Channel Models in Closed Form

  ArXiv.org · 2026-04-03

  articleOpen access1st authorCorresponding

  Dynamic fading channels are modeled at two fundamentally different levels of abstraction. At the physical layer, the standard representation is a correlated Gaussian process, such as the dB-domain signal power in log-normal shadow fading. At the link layer, the dominant abstraction is the Gilbert-Elliott (GE) two-state Markov chain, which compresses the channel into a binary ``decodable or not'' sequence with temporal memory. Both models are ubiquitous, yet practitioners who need GE parameters from an underlying Gaussian fading model must typically simulate the mapping or invoke continuous-time level-crossing approximations that do not yield discrete-slot transition probabilities in closed form. This paper provides an exact, closed-form bridge. By thresholding the Gaussian process at discrete slot boundaries, we derive the GE transition probabilities via Owen's $T$-function for any threshold, reducing to an elementary arcsine identity when the threshold equals the mean. The formulas depend on the covariance kernel only through the one-step correlation coefficient $ρ= K(D)/K(0)$, making them applicable to any stationary Gaussian fading model. The bridge reveals how kernel smoothness governs the resulting link-layer dynamics: the GE persistence time grows linearly in the correlation length $T_c$ for a smooth (squared-exponential) kernel but only as $\sqrt{T_c}$ for a rough (exponential/Ornstein--Uhlenbeck) kernel. We further quantify when the first-order GE chain is a faithful approximation of the full binary process and when it is not, reconciling two diagnostics, the one-step Markov gap and the run-length total-variation distance, that can trend in opposite directions. Monte Carlo simulations validate all theoretical predictions.

  PublisherOA PDF

• SATORIS-N: Spectral Analysis based Traffic Observation Recovery via Informed Subspaces and Nuclear-norm minimization

  Open MIND · 2026-02-03

  preprintSenior author

  Traffic-density matrices from different days exhibit both low rank and stable correlations in their singular-vector subspaces. Leveraging this, we introduce SATORIS-N, a framework for imputing partially observed traffic-density by informed subspace priors from neighboring days. Our contribution is a subspace-aware semidefinite programming (SDP)} formulation of nuclear norm that explicitly informs the reconstruction with prior singular-subspace information. This convex formulation jointly enforces low rank and subspace alignment, providing a single global optimum and substantially improving accuracy under medium and high occlusion. We also study a lightweight implicit subspace-alignment} strategy in which matrices from consecutive days are concatenated to encourage alignment of spatial or temporal singular directions. Although this heuristic offers modest gains when missing rates are low, the explicit SDP approach is markedly more robust when large fractions of entries are missing. Across two real-world datasets (Beijing and Shanghai), SATORIS-N consistently outperforms standard matrix-completion methods such as SoftImpute, IterativeSVD, statistical, and even deep learning baselines at high occlusion levels. The framework generalizes to other spatiotemporal settings in which singular subspaces evolve slowly over time. In the context of intelligent vehicles and vehicle-to-everything (V2X) systems, accurate traffic-density reconstruction enables critical applications including cooperative perception, predictive routing, and vehicle-to-infrastructure (V2I) communication optimization. When infrastructure sensors or vehicle-reported observations are incomplete - due to communication dropouts, sensor occlusions, or sparse connected vehicle penetration-reliable imputation becomes essential for safe and efficient autonomous navigation.

  DOI

• From Gaussian Fading to Gilbert-Elliott: Bridging Physical and Link-Layer Channel Models in Closed Form

  arXiv (Cornell University) · 2026-04-03

  preprintOpen access1st authorCorresponding

  Dynamic fading channels are modeled at two fundamentally different levels of abstraction. At the physical layer, the standard representation is a correlated Gaussian process, such as the dB-domain signal power in log-normal shadow fading. At the link layer, the dominant abstraction is the Gilbert-Elliott (GE) two-state Markov chain, which compresses the channel into a binary ``decodable or not'' sequence with temporal memory. Both models are ubiquitous, yet practitioners who need GE parameters from an underlying Gaussian fading model must typically simulate the mapping or invoke continuous-time level-crossing approximations that do not yield discrete-slot transition probabilities in closed form. This paper provides an exact, closed-form bridge. By thresholding the Gaussian process at discrete slot boundaries, we derive the GE transition probabilities via Owen's $T$-function for any threshold, reducing to an elementary arcsine identity when the threshold equals the mean. The formulas depend on the covariance kernel only through the one-step correlation coefficient $ρ= K(D)/K(0)$, making them applicable to any stationary Gaussian fading model. The bridge reveals how kernel smoothness governs the resulting link-layer dynamics: the GE persistence time grows linearly in the correlation length $T_c$ for a smooth (squared-exponential) kernel but only as $\sqrt{T_c}$ for a rough (exponential/Ornstein--Uhlenbeck) kernel. We further quantify when the first-order GE chain is a faithful approximation of the full binary process and when it is not, reconciling two diagnostics, the one-step Markov gap and the run-length total-variation distance, that can trend in opposite directions. Monte Carlo simulations validate all theoretical predictions.

  PublisherDOI

• RISE: Resource-Efficient RFID Security Protocol in IoT with S-Box and Elliptic Curve Cryptography

  IEEE Internet of Things Journal · 2026-01-01

  articleSenior author

  Ensuring the security and privacy of RFID systems is paramount to prevent unauthorized access and data breaches. In addition, addressing the insecurity of public keys in RFID tags and preventing server-side tag distinction requires a multifaceted approach. This paper proposes an RFID-enabled protocol, called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RISE</i>, Resource-Efficient RFID Security Protocol in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">I</i>oT with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</i>-Box and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">E</i>lliptic Curve Cryptography, that integrates S-box and elliptic curve cryptography (ECC) to enhance the security of RFID communication. The protocol leverages S-boxes to introduce non-linearity and confusion, thereby thwarting potential attacks such as cryptanalysis and replay attacks. Additionally, ECC is employed for key exchange and authentication, leveraging computational efficiency and strong security properties. Formal proofs and analysis results demonstrate that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RISE</i> provides robust security and can effectively resist various types of attacks. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RISE</i> presents notable improvements, with an average 51.18% reduction in computational overhead compared to similar protocols. This advancement not only boosts performance but also enhances resource efficiency.

  PublisherDOI

• SATORIS-N: Spectral Analysis based Traffic Observation Recovery via Informed Subspaces and Nuclear-norm minimization

  ArXiv.org · 2026-02-03

  articleOpen accessSenior author

  Traffic-density matrices from different days exhibit both low rank and stable correlations in their singular-vector subspaces. Leveraging this, we introduce SATORIS-N, a framework for imputing partially observed traffic-density by informed subspace priors from neighboring days. Our contribution is a subspace-aware semidefinite programming (SDP)} formulation of nuclear norm that explicitly informs the reconstruction with prior singular-subspace information. This convex formulation jointly enforces low rank and subspace alignment, providing a single global optimum and substantially improving accuracy under medium and high occlusion. We also study a lightweight implicit subspace-alignment} strategy in which matrices from consecutive days are concatenated to encourage alignment of spatial or temporal singular directions. Although this heuristic offers modest gains when missing rates are low, the explicit SDP approach is markedly more robust when large fractions of entries are missing. Across two real-world datasets (Beijing and Shanghai), SATORIS-N consistently outperforms standard matrix-completion methods such as SoftImpute, IterativeSVD, statistical, and even deep learning baselines at high occlusion levels. The framework generalizes to other spatiotemporal settings in which singular subspaces evolve slowly over time. In the context of intelligent vehicles and vehicle-to-everything (V2X) systems, accurate traffic-density reconstruction enables critical applications including cooperative perception, predictive routing, and vehicle-to-infrastructure (V2I) communication optimization. When infrastructure sensors or vehicle-reported observations are incomplete - due to communication dropouts, sensor occlusions, or sparse connected vehicle penetration-reliable imputation becomes essential for safe and efficient autonomous navigation.

  PublisherOA PDF

• Target-Rate Least-Squares Power Allocation over Parallel Channels

  ArXiv.org · 2026-03-06

  articleOpen access1st authorCorresponding

  We study power allocation over $N$ parallel Gaussian channels, such as OFDM subcarriers, when each channel has a desired target spectral efficiency. Given channel gain-to-noise coefficients $a_i>0$ and per-channel targets $T_i\ge 0$, we minimize the total squared rate deviation $\sum_{i=1}^{N}(\log_2(1+a_iP_i)-T_i)^2$ subject to a sum-power constraint $\sum_i P_i \le P_{\mathrm{tot}}$ and nonnegativity $P_i \ge 0$. We prove that the optimal allocation never overshoots any target and may leave power unused when all targets are jointly feasible, a structure fundamentally different from classical waterfilling. Using the KKT conditions, we derive a per-channel closed-form solution in terms of the Lambert~W function on the active set and reduce the remaining computation to a one-dimensional monotone bisection for the dual variable. The resulting algorithm runs in $O(N\log(1/\varepsilon))$ time and achieves up to 1{,}890$\times$ speedup over general-purpose numerical solvers at $N=1024$ channels. Numerical experiments over Rayleigh fading channels confirm that the closed-form solution matches numerical optimization to machine precision and demonstrate superior target-tracking performance compared to waterfilling, uniform allocation, and proportional fairness across a range of operating conditions.

  PublisherOA PDF

• Crowd-SFT: Crowdsourcing for LLM Alignment

  2025-07-21 · 1 citations

  articleSenior author

  Large Language Models (LLMs) rely on Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF) to align with human preferences. Yet both approaches typically depend on small, costly, and potentially biased annotator pools. We introduce a scalable, crowd-sourced SFT framework that enables broader participation without intensive annotator training. Our method leverages iterative, competitive model updates and a point-based reward mechanism correlated with Shapley values to ensure fair attribution of contributions. Simulations show that our multi-model selection process improves convergence by up to 55% over traditional SFT. Additionally, our point-based system closely tracks individual impact and scales effectively across diverse grouping and evaluation strategies. This work offers a practical, incentive-aligned path for decentralized LLM alignment.

  PublisherDOI

• Crowd-SFT: Crowdsourcing for LLM Alignment

  ArXiv.org · 2025-06-04

  preprintOpen accessSenior author

  Large Language Models (LLMs) increasingly rely on Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF) to align model responses with human preferences. While RLHF employs a reinforcement learning approach with a separate reward model, SFT uses human-curated datasets for supervised learning. Both approaches traditionally depend on small, vetted groups of annotators, making them costly, prone to bias, and limited in scalability. We propose an open, crowd-sourced fine-tuning framework that addresses these limitations by enabling broader feedback collection for SFT without extensive annotator training. Our framework promotes incentive fairness via a point-based reward system correlated with Shapley values and guides model convergence through iterative model updates. Our multi-model selection framework demonstrates up to a 55% reduction in target distance over single-model selection, enabling subsequent experiments that validate our point-based reward mechanism's close alignment with Shapley values (a well-established method for attributing individual contributions) thereby supporting fair and scalable participation.

  PublisherOA PDFDOI

• Computing the Saturation Throughput for Heterogeneous p-CSMA in a General Wireless Network

  2025-08-04

  articleOpen accessSenior author

  A well-known expression for the saturation throughput of heterogeneous transmitting nodes in a wireless network using p-CSMA, derived from Renewal Theory, implicitly assumes that all transmitting nodes are in range of, and therefore conflicting with, each other. This expression, as well as simple modifications of it, does not correctly capture the saturation throughput values when an arbitrary topology is specified for the conflict graph between transmitting links. For example, we show numerically that calculations based on renewal theory can underestimate throughput by 48-62% for large packet sizes when the conflict graph is represented by a star topology. This is problematic because real-world wireless networks, such as wireless IoT mesh networks, are often deployed over a large area, resulting in non-complete conflict graphs. To address this gap, we present a computational approach based on a novel Markov chain formulation that yields the exact saturation throughput for each node in the general network case for any given set of access probabilities, as well as a more compact expression for the special case where the packet length is twice the slot length. Using our approach, we show how the transmit probabilities could be optimized to maximize weighted utility functions of the saturation throughput values. This would allow a wireless system designer to set transmit probabilities to achieve desired throughput trade-offs in any given deployment.

  PublisherOA PDFDOI

Recent grants

• NeTS: Small: GOALI: Information Centric Networking on Wheels (IC NoW) - Architecture and Protocols

  NSF · $350k · 2012–2016

• ECCS - EARS: Collaborative Research: Enhanced Radio Spectrum via Information Acquisition and Learning

  NSF · $250k · 2012–2016

• EAGER: Towards an Open Source Backpressure Protocol Stack for Wireless Networks

  NSF · $240k · 2010–2013

• CAREER: Mathematical Models for Querying and Routing in Wireless Sensor Networks

  NSF · $420k · 2004–2009

• NeTS-NOSS: Data-Centric Active Querying in Sensor Networks

  NSF · $750k · 2004–2008

Frequent coauthors

• Tarek Abdelzaher

  192 shared

• Viktor K. Prasanna

  136 shared

• Jie Gao

  130 shared

• Sotiris Nikoletseas

  130 shared

• Magnús M. Halldórsson

  Reykjavík University

  129 shared

• Luca Mottola

  Politecnico di Milano

  128 shared

• Arvin Hekmati

  University of Southern California

  85 shared

• Eugenio Grippo

  University of Southern California

  74 shared

Education

• Ph.D., Electrical Engineering

  University of California, Los Angeles

  1995

• M.S., Electrical Engineering

  University of California, Los Angeles

  1991

• B.S., Electrical Engineering

  Indian Institute of Technology, Madras

  1988

Awards & honors

• IEEE Fellow (2023)
• Popular Science Brilliant 10 (2015)
• TR-35, MIT Technology Review's Top 35 Innovators Under 35 (2…
• IEEE - Eta Kappa Nu Outstanding Young Electrical and Compute…
• ASEE Frederick E. Terman Award (2010)