arXiv subjects

Saurabh Deochake

Publications and source records attributed to Saurabh Deochake.

13 recordsLinked to original sources

arXiv2026

ToolGuardian: Declarative Security for AI Agent-Tool Interactions

LLM agents increasingly rely on external tools, expanding capability while creating a new security boundary: third-party tools may appear benign at the interface level while embedding unsafe behavior in implementation. Existing defenses rely on weak metadata, collapse characterization and policy judgment into a single decision, or use heuristic/LLM enforcement that lacks deterministic, auditable reasoning over task context and multi-tool composition. This paper presents ToolGuardian, a policy-driven framework for securing agent-tool interactions through pre-admission vetting and task-aware runtime authorization. ToolGuardian uses progressive characterization to convert evidence into structured facts: descriptions capture declared intent, system-call traces expose coarse behavior, mock execution reveals observed effects, and source analysis identifies latent behavior. ToolGuardian's core contribution is an Answer Set Programming (ASP)-based declarative policy layer that reasons explicitly over capabilities, effects, task context, and composition. We compare ASP against heuristic and LLM-based policy realizations using identical inputs and output contracts. We evaluate ToolGuardian on 16 MCP-style tools, including 8 malicious variants derived from real open-source tools, and 20 runtime scenarios. For vetting, ASP reaches a deny-class F1 of 0.86 and 88% accuracy using description, syscall, and observed-effect evidence. For runtime authorization, fully specified realizations classify all scenarios correctly, while ablations show that removing compositional and conformance rules substantially degrades performance.

cs.CR↗

arXiv2026

Heartbeat-Bound Hierarchical Credentials: Cryptographic Revocation for AI Agent Swarms

Autonomous AI agents that spawn sub-agent swarms create a safety gap: existing credential revocation mechanisms, OAuth~2.0 introspection, OCSP, and W3C Status Lists, require network connectivity to a central authority, leaving ``zombie agents'' executing privileged operations for minutes to hours after operator shutdown. We present Heartbeat-Bound Hierarchical Credentials (HBHC), a cryptographic protocol that binds credential validity to periodic parent liveness proofs. Verifiers enforce freshness using only a cached public key and local clock; no network round-trip is required. When heartbeat generation ceases, all descendant credentials become unusable within a deterministically bounded window $W_z \le W_{\max} + \Delta_h + \epsilon$, conditional on bounded clock skew and parent keys held in secure enclaves. Evaluation at the protocol layer and with real LLM-backed agent swarms (GPT-4o-mini) demonstrates a 90$\times$ reduction in the zombie window over OAuth~2.0, 0.26~ms full authentication in Rust, 18,000+ verifications per second under concurrent HTTP load, and stable per-verification latency from 10 to 10,000 agents. Real-agent experiments show 0.71\% end-to-end overhead on tool calls, zero post-revocation tool calls under prompt injection that bypasses application-layer guardrails, and cascading revocation across a 49-agent four-level hierarchy within the theoretical bound.

cs.CR↗

arXiv2025

Cost Trade-offs of Reasoning and Non-Reasoning Large Language Models in Text-to-SQL

While Text-to-SQL systems achieve high accuracy, existing efficiency metrics like the Valid Efficiency Score prioritize execution time, a metric we show is fundamentally decoupled from consumption-based cloud billing. This paper evaluates cloud query execution cost trade-offs between reasoning and non-reasoning Large Language Models by performing 180 Text-to-SQL query executions across six LLMs on Google BigQuery using the 230 GB StackOverflow dataset. Our analysis reveals that reasoning models process 44.5% fewer bytes than non-reasoning counterparts while maintaining equivalent correctness at 96.7% to 100%, and that execution time correlates weakly with query cost at $r=0.16$, indicating that speed optimization does not imply cost efficiency. Non-reasoning models also exhibit extreme cost variance of up to 3.4$\times$, producing outliers exceeding 36 GB per query, over 20$\times$ the best model's 1.8 GB average, due to missing partition filters and inefficient joins. We identify these prevalent inefficiency patterns and provide deployment guidelines to mitigate financial risks in cost-sensitive enterprise environments.

cs.DB↗

arXiv2025

A Multi-Cloud Framework for Zero-Trust Workload Authentication

Static, long-lived credentials for workload authentication create untenable security risks that violate Zero-Trust principles. This paper presents a multi-cloud framework using Workload Identity Federation (WIF) and OpenID Connect (OIDC) for secretless authentication. Our approach uses cryptographically-verified, ephemeral tokens, allowing workloads to authenticate without persistent private keys and mitigating credential theft. We validate this framework in an enterprise-scale Kubernetes environment, which significantly reduces the attack surface. The model offers a unified solution to manage workload identities across disparate clouds, enabling future implementation of robust, attribute-based access control.

cs.CR↗

arXiv2024

ABACUS: A FinOps Service for Cloud Cost Optimization

In recent years, as more enterprises have moved their infrastructure to the cloud, significant challenges have emerged in achieving holistic cloud spend visibility and cost optimization. FinOps practices provide a way for enterprises to achieve these business goals by optimizing cloud costs and bringing accountability to cloud spend. This paper presents ABACUS - Automated Budget Analysis and Cloud Usage Surveillance, a FinOps solution for optimizing cloud costs by setting budgets, enforcing those budgets through blocking new deployments, and alerting appropriate teams if spending breaches a budget threshold. ABACUS also leverages best practices like Infrastructure-as-Code to alert engineering teams of the expected cost of deployment before resources are deployed in the cloud. Finally, future research directions are proposed to advance the state of the art in this important field.

cs.DC↗

arXiv2023

Cloud and AI Infrastructure Cost Optimization: A Comprehensive Review of Strategies and Case Studies

Cloud computing has revolutionized the way organizations manage their IT infrastructure, but it has also introduced new challenges, such as managing cloud costs. The rapid adoption of artificial intelligence (AI) and machine learning (ML) workloads has further amplified these challenges, with GPU compute now representing 40-60\% of technical budgets for AI-focused organizations. This paper provides a comprehensive review of cloud and AI infrastructure cost optimization techniques, covering traditional cloud pricing models, resource allocation strategies, and emerging approaches for managing AI/ML workloads. We examine the dramatic cost reductions in large language model (LLM) inference which has decreased by approximately 10x annually since 2021 and explore techniques such as model quantization, GPU instance selection, and inference optimization. Real-world case studies from Amazon Prime Video, Pinterest, Cloudflare, and Netflix showcase practical application of these techniques. Our analysis reveals that organizations can achieve 50-90% cost savings through strategic optimization approaches. Future research directions in automated optimization, sustainability, and AI-specific cost management are proposed to advance the state of the art in this rapidly evolving field.

cs.DC↗

arXiv2023

Comparative Study of Virtual Machines and Containers for DevOps Developers

This paper presents a comparative study of virtual machines (VMs) and containers for DevOps developers. The study explores the benefits and drawbacks of each technology in terms of their functionality, performance, security, and resource utilization. The paper examines the underlying architecture of VMs and containers, and how they differ from each other. The study includes a series of experiments that compare the performance and resource utilization of VMs and containers in different scenarios. The experiments evaluate factors such as startup time, memory usage, disk I/O, network latency, scalability, and administrative overhead. Finally, the paper provides recommendations for DevOps developers on which technology to choose based on their specific requirements and use cases. Overall, the study aims to provide a comprehensive understanding of the strengths and limitations of VMs and containers, helping developers to make informed decisions when choosing between them.

cs.OS↗

arXiv2022

Belief-Desire-Intention (BDI) Multi-agent System for Cloud Marketplace Negotiation

With the evolution of cloud computing, there has been a rise of large enterprises extending their infrastructure and workloads into the public cloud. This paper proposes a full-fledged framework for a Belief-Desire-Intention (BDI) multi-agent-based cloud marketplace system for cloud resources. Each party in the cloud marketplace system supports a BDI agent for autonomous decision making and negotiation to facilitate automated buying and selling of resources. Additionally, multiple BDI agents from an enterprise competing for the same cloud resource can consult with each other via Master Negotiation Clearing House to minimize the overall cost function for the enterprise while negotiating for a cloud resource. The cloud marketplace system is further augmented with assignments of behavior norm and reputation index to the agents to facilitate trust among them.

cs.MA↗

arXiv2022

Identity and Access Management Framework for Multi-tenant Resources in Hybrid Cloud Computing

While more organizations have been trying to move their infrastructure to the cloud in recent years, there have been significant challenges in how identities and access are managed in a hybrid cloud setting. This paper showcases a novel identity and access management framework for shared resources in a multi-tenant hybrid cloud environment. The paper demonstrates a method to implement the "mirror" identities of on-premise identities in the cloud. Following the best security practices, the framework ensures that only rightful users can use their mirror identities in the cloud. Furthermore, the paper also proposes a technique in scaling the framework to accommodate large-scale enterprises. The framework exhibited in the paper provides a comprehensive and scalable solution for enterprises to implement identity and access control in their hybrid cloud infrastructure. Although the paper focuses on implementing the framework in Google Cloud Platform, it can be easily applied to any major public cloud platform.

cs.DC↗

arXiv2022

BigBird: Big Data Storage and Analytics at Scale in Hybrid Cloud

Implementing big data storage at scale is a complex and arduous task that requires an advanced infrastructure. With the rise of public cloud computing, various big data management services can be readily leveraged. As a critical part of Twitter's "Project Partly Cloudy", the cold storage data and analytics systems are being moved to the public cloud. This paper showcases our approach in designing a scalable big data storage and analytics management framework using BigQuery in Google Cloud Platform while ensuring security, privacy, and data protection. The paper also discusses the limitations on the public cloud resources and how they can be effectively overcome when designing a big data storage and analytics solution at scale. Although the paper discusses the framework implementation in Google Cloud Platform, it can easily be applied to all major cloud providers.

cs.DC↗

arXiv2022

HENRI: High Efficiency Negotiation-based Robust Interface for Multi-party Multi-issue Negotiation over the Internet

This paper proposes a framework for a full fledged negotiation system that allows multi party multi issue negotiation. It focuses on the negotiation protocol to be observed and provides a platform for concurrent and independent negotiation on individual issues using the concept of multi threading. It depicts the architecture of an agent detailing its components. The paper sets forth a hierarchical pattern for the multiple issues concerning every party. The system also provides enhancements such as the time-to-live counters for every advertisement, refinement of utility considering non-functional attributes, prioritization of issues, by assigning weights to issues.

cs.AI↗

arXiv2020

An Agent-based Cloud Service Negotiation in Hybrid Cloud Computing

With the advent of evolution of cloud computing, large organizations have been scaling the on-premise IT infrastructure to the cloud. Although this being a popular practice, it lacks comprehensive efforts to study the aspects of automated negotiation of resources among cloud customers and providers. This paper proposes a full-fledged framework for the multi-party, multi-issue negotiation system for cloud resources. It introduces a robust cloud marketplace system to buy and sell cloud resources. The Belief-Desire-Intention (BDI) model-based cloud customer and provider agents concurrently negotiate on multiple issues, pursuing a hybrid tactic of time and resource-based dynamic deadline algorithms to generate offers and counter-offers. The cloud marketplace-based system is further augmented with the assignment of behavior norm score and reputation index to the agents to establish trust among them.

cs.MA↗

arXiv2012

MAINWAVE: Multi Agents and Issues Negotiation for Web using Alliance Virtual Engine

This paper showcases an improved architecture for a complete negotiation system that permits multi party multi issue negotiation. The concepts of multithreading and concurrency has been utilized to perform parallel execution. The negotiation history has been implemented that stores all the records of the messages exchanged for every successful and rejected negotiation process and implements the concepts of artificial intelligence in determination of proper weights for a valid negotiation mechanism. The issues are arranged in a hierarchical pattern so as to simplify the representation and priorities are assigned to each issue, which amounts to its relative importance. There is refinement of utilities by consideration of the non-functional attributes. So as to avoid overloading of the system, a maximum number of parties are allowed to participate in the entire mechanism and if more parties arrive, they're put into a waiting queue in accordance to certain criteria such as the first come first serve or the relative priorities. This helps in fault tolerance. It also supports the formation of alliances among the various parties while carrying out a negotiation.

cs.MA↗