2026· Poster Volume 0008 The 2026 Twenty-Second International Conference on Intelligent Computing July 23-26, 2026 Toronto, Canada· pp. 702-710· 0 citations
TL;DR
This paper organizes the area into a structured taxonomy along five axes: the de-tection capability targeted, the analysis paradigm employed, the agent archi-tecture, the degree of autonomy, and the evaluation methodology.
Abstract
Software vulnerabilities have become a recognized national-security risk, yet the volume of code and the sophistication of threats now far outpace what manual security review and the limited supply of expert security engineers can sustain. A new class of systems has emerged in response: agentic systems for software security detection, which couple large language models with planning, memory, and external tools so that they can autonomously analyze codebases, reason about program behavior, and identify, triage, and help re-mediate vulnerabilities. The field has grown rapidly but unevenly, and its terminology, capabilities, and evaluation practices remain fragmented. This paper organizes the area into a structured taxonomy along five axes: the de-tection capability targeted, the analysis paradigm employed, the agent archi-tecture, the degree of autonomy, and the evaluation methodology. We popu-late the taxonomy with representative systems, including the cyber reasoning systems demonstrated at the DARPA AI Cyber Challenge, and we use it to compare designs, surface recurring patterns, and expose gaps. We find that the strongest results combine learned reasoning with classical program analy-sis and tool use rather than relying on either alone, and that repository-scale detection, trustworthy triage, and reproducible evaluation remain the principal open challenges. The taxonomy is intended as a shared vocabulary and a roadmap for building the next generation of autonomous software-security systems.
Cyber-capable AI agents combine language models with tools, memory, and execution environments to perform multi-step offensive-security tasks. Existing work separately measures cyber capability and catalogs attacks against agent components, but provides less guidance on containing a capable agent within the environments used to evaluate it. This review synthesizes five vulnerability classes at that boundary: multi-step offensive chains, objectives that conflict with sandbox boundaries, supply-chain and credential exposure, persistent command-and-control, and the speed of automated action. We use two separate preliminary incident records: the reported July 2026 Hugging Face/OpenAI evaluation breach and Anthropic's subsequent three-incident evaluation review. A comparative evidence protocol distinguishes record-specific factual claims from the shared systems lesson: the evaluation environment is itself part of the security boundary. Across the taxonomy and records, we examine controls for containment, privilege separation, provenance, and responder access, including the dual-use problem that defensive artifacts may also enable misuse. The review identifies practical priorities for evaluating cyber capability together with the security of the environment in which that capability is exercised.
A structured taxonomy is proposed to organize various dimensions of AI-driven cybersecurity; review them critically; and finally, discuss key challenges, open problems, and emerging trends.
The rapid enterprise adoption of agentic artificial intelligence (AI) has introduced a category of security risk that existing cyber security frameworks were not designed to address. With 78 per cent of Fortune 500 companies projected to deploy agentic AI by 2026 and the global market expected to reach US$89.6bn, the attack surface created by these autonomous workflows demands urgent attention from security practitioners. This paper examines the distinct threat model presented by agentic AI, drawing on recent high-profile incidents, including the weaponisation of a large language model in a state-sponsored espionage campaign affecting 30 organisations and the compromise of an open-source agent framework exposing 30,000 Internet-facing instances, to illustrate the consequences of inadequate controls. Grounded in the Open Worldwide Application Security Project’s Top 10 for Agentic Applications (2026) and the National Institute of Standards and Technology’s ongoing agentic AI security initiative, the paper proposes a five-layer defence-in-depth framework encompassing input validation, identity and least privilege, runtime sandboxing, human-in-the-loop governance, and continuous behavioural monitoring. It identifies sandboxing and least-privilege enforcement as the highest return on investment controls, provides a prioritised implementation roadmap, and discusses the emerging paradigm of cryptographic workflow authentication. The analysis concludes that organisations treating agentic AI security as an extension of traditional application security will find themselves critically exposed and that a purpose-built security architecture is now a business imperative. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.
Sushma Mahadevaswamy· International Conference on...· 0 citations
The findings suggest that combining open-source SIEM, workflow automation, and LLM-based reasoning with human supervision offers a practical, low-cost, and reliable approach for strengthening incident response capability in resource-constrained environments.
As software systems continue to evolve in complexity, ensuring their security has become an increasingly critical challenge. Traditional security approaches often fail to adapt to dynamic threat landscapes and sophisticated cyber-attacks. This paper presents the design and development of an Artificial Intelligence (AI)-driven Knowledge Processing System (KPS) aimed at optimizing the security of software systems. The proposed system leverages AI techniques such as machine learning, natural language processing, and expert systems to analyze threat patterns, detect anomalies, and suggest real-time mitigation strategies. By integrating continuous learning from security data and contextual knowledge, the system enhances decision-making and predictive capabilities for threat prevention and response. This research highlights system architecture, implementation methodology, and experimental validation to demonstrate the system’s efficacy. Results show significant improvement in threat detection accuracy, response time, and overall system resilience, suggesting that AI-based KPS can be a powerful tool in the software security domain.
Blessling Nova, Tunde Ogunleye· International Journal of Dat...· 0 citations
There is a structural shift in software engineering that has never been seen before. Deterministic, human-written source code is giving way to dynamic computing environments powered by artificial intelligence that can program itself. These multi-agent architectures create, assemble, and run real-time operating environments on their own. This fluid movement exposes computer infrastructure to systemic, deep architectural vulnerabilities even while it promises historic improvements in hardware flexibility and execution speed. Software contexts that dynamically rewrite their own execution logic cannot be secured using traditional cybersecurity techniques, particularly static analysis, signature detection, and perimeter protection.
The vulnerabilities present in unsupervised automated system compilation are thoroughly examined in this research. We model target threat vectors, such as the compression of zero-day discovery-to-exploit lifecycles by adversarial models, the amplification of design vulnerabilities during unsupervised generation, and excessive agency across multi-layered execution environments. In order to show that using secondary reasoning models for auditing results in common logical blind spots and systemic failure routes, we explicitly analyze the cognitive recursive loop dilemma ("Who Watches the Watcher?" conundrum). We designate automated mathematical validation as the ultimate gatekeeper for dynamic system compilation in order to overcome this structural constraint. We demonstrate the specific dangers of unverified synthesis using recent real-world case studies including flaws in autonomous developer interfaces and automated schema deployment tools. In order to safeguard the upcoming generation of computing runtimes, we finally suggest an operational blueprint for an Autonomous Defensive Layer (ADL) that enforces tight micro-virtualization, continuous shadow reasoning execution, and deterministic formal verification pipelines.
Muhammad Anjum· International Journal of Dig...· 0 citations