arXiv subjects

Basavesh Ammanaghatta Shivakumar

Publications and source records attributed to Basavesh Ammanaghatta Shivakumar.

4 recordsLinked to original sources

arXiv2026

AgentFlow: A Flow-Centric Policy Language and Framework for Securing LLM Agent Systems

LLM agents increasingly read untrusted content, invoke external tools, access private data, and delegate work to other agents. Harm often arises not from a single unsafe action but from the flow of sensitive data across a sequence of otherwise plausible steps. We present AgentFlow, a flow-centric policy language and runtime enforcement model for specifying where data may travel in agent systems. Policies are defined over labeled runtime edges and constrain which tools may receive sensitive fields, which sinks may receive released data, and what authority may cross delegation boundaries. The language supports flow and path rules, task-scoped capabilities, controlled release, and stateful taint semantics. A runtime reference monitor mediates agent actions, and a bounded SMT-based verifier checks safety properties for a structured policy fragment. We evaluate AgentFlow on multiple agent benchmarks. In our prototype, seven safety properties verify in under 0.5 seconds each, and the verifier catches all seeded unsafe policy variants in our study. On 949 AgentDojo injected cases across four suites, AgentFlow reduces confirmed compromise from 33.0\% to 0.0\% while improving aggregate utility from 46.7\% to 63.3\%. On a 200-case AgentDyn Dailylife benchmark, it reduces confirmed compromise from 73.5\% to 0.0\% while preserving near-baseline utility (44.5\% to 43.5\%). Breadth checks across ASB, InjecAgent, BIPIA, AgentHarm, and MCPTox replays suggest that the configured policies block the benchmark-specified policy-visible attacker flows; in ASB's direct-prompt-injection harness, attack success is 0/1{,}200. These results are preliminary and scoped to the modeled policy-visible agent behaviors and evaluated benchmarks.

cs.CR↗

arXiv2024

HMTRace: Hardware-Assisted Memory-Tagging based Dynamic Data Race Detection

Data race, a category of insidious software concurrency bugs, is often challenging and resource-intensive to detect and debug. Existing dynamic race detection tools incur significant execution time and memory overhead while exhibiting high false positives. This paper proposes HMTRace, a novel Armv8.5-A memory tag extension (MTE) based dynamic data race detection framework, emphasizing low compute and memory requirements while maintaining high accuracy and precision. HMTRace supports race detection in userspace OpenMP- and Pthread-based multi-threaded C applications. HMTRace showcases a combined f1-score of 0.86 while incurring a mean execution time overhead of 4.01% and peak memory (RSS) overhead of 54.31%. HMTRace also does not report false positives, asserting all reported races.

cs.DC↗

arXiv2023

Robust Constant-Time Cryptography

The constant-time property is considered the security standard for cryptographic code. Code following the constant-time discipline is free from secret-dependent branches and memory accesses, and thus avoids leaking secrets through cache and timing side-channels. The constant-time property makes a number of implicit assumptions that are fundamentally at odds with the reality of cryptographic code. Constant-time is not robust. The first issue with constant-time is that it is a whole-program property: It relies on the entirety of the code base being constant-time. But, cryptographic developers do not generally write whole programs; rather, they provide libraries and specific algorithms for other application developers to use. As such, developers of security libraries must maintain their security guarantees even when their code is operating within (potentially untrusted) application contexts. Constant-time requires memory safety. The whole-program nature of constant-time also leads to a second issue: constant-time requires memory safety of all the running code. Any memory safety bugs, whether in the library or the application, will wend their way back to side-channel leaks of secrets if not direct disclosure. And although cryptographic libraries should (and are) written to be memory-safe, it is unfortunately unrealistic to expect the same from every application that uses each library. We formalize robust constant-time and build a RobustIsoCrypt compiler that transforms the library code and protects the secrets even when they are linked with untrusted code. Our evaluation with SUPERCOP benchmarking framework shows that the performance overhead is less than five percent on average.

cs.CR↗

arXiv2020

On the Feasibility of Exploiting Traffic Collision Avoidance System Vulnerabilities

Traffic Collision Avoidance Systems (TCAS) are safety-critical systems required on most commercial aircrafts in service today. However, TCAS was not designed to account for malicious actors. While in the past it may have been infeasible for an attacker to craft radio signals to mimic TCAS signals, attackers today have access to open-source digital signal processing software, like GNU Radio, and inexpensive software defined radios (SDR) that enable the transmission of spurious TCAS messages. In this paper, methods, both qualitative and quantitative, for analyzing TCAS from an adversarial perspective are presented. To demonstrate the feasibility of inducing near mid-air collisions between current day TCAS-equipped aircraft, an experimental Phantom Aircraft generator is developed using GNU Radio and an SDR against a realistic threat model.

eess.SP↗