From AI to Supply Chain: Why Attack Surface Reduction is No Longer Optional

Presentation

Presentation Overview

C3-12 | Oct 23 (Thu) 18:30-19:30 [Simultaneous Translation] Night Session with Light Refreshments!

Note: This session was delivered in English with simultaneous translation services provided.

AI is lowering the barriers for attackers, supply chains are becoming more complex with invisible dependencies, and security incidents are increasingly bypassing traditional defenses by exploiting misconfigurations and vulnerabilities in abstraction layers.

This presentation connects the dots of recent cyber trends, from LLM-powered reconnaissance to CI/CD abuse and cloud-based over-exposure, making a compelling case for attack surface reduction as a fundamental security principle rather than an afterthought.

In a world where threats automatically expand and risks lurk in the shadows, reducing exposure may be the most effective countermeasure we can implement.

Speaker Profile

Anant Shrivastava | Cyfinoid Research

Anant Shrivastava is the founder of Cyfinoid Research. He has extensive experience in security (both offensive and defensive), development, and operations.

He has a rich track record as a trainer and speaker at prestigious conferences including Black Hat (USA, Asia, EU), Nullcon, and c0c0n. He also leads the open source project CodeVigilant and curates the Hacking Archives of India.

Outside of his official work, he contributes to open communities such as null community, Garage4Hackers, hasgeek, and OWASP, all with the common goal of spreading information security knowledge.

Event Information

Event: Security Days 2025 Fall
Venue: Tokyo Venue
Date & Time: October 23, 2025 (Thu) 18:30-19:30
Format: Night Session (with light refreshments)
Translation: Simultaneous translation available
Official Site: Security Days 2025 Fall

日本語版 / Japanese Version

講演概要

C3-12｜10.23(木) 18:30-19:30【同時通訳あり】軽食付きのナイトセッション！

注：このセッションは英語で実施され、同時通訳サービスが提供されました。

AIは攻撃者の障壁を下げ、サプライチェーンは目に見えない依存関係で複雑化し、セキュリティインシデントは誤設定や抽象化レイヤーの脆弱性を突くことで従来の防御をすり抜けることが増えている。

本講演では、LLMを活用したリコンからCI/CDの乱用、クラウドベースの過剰な露出まで、近年のサイバー・トレンドを点と点で結び付け、攻撃対象領域の削減を後回しにするのではなく、セキュリティの基本原則とする説得力のあるケースを紹介する。

脅威が自動的に拡大し、リスクが見え隠れする世界において、露出を減らすことは私たちにできる最も効果的な対策かもしれない。

講演者プロフィール

Anant Shrivastava | Cyfinoid Research

Anant Shrivastavaは、Cyfinoid Researchの創設者である。セキュリティ（攻撃および防御）、開発、運用の分野で経験を積む。

Black Hat（USA, アジア, EU）、Nullcon、c0c0nをはじめとする著名なカンファレンスで、トレーナーおよびスピーカーとして豊富な実績を持つ。また、オープンソースプロジェクトのCodeVigilantを主導し、Hacking Archives of Indiaをキュレーションしている。

公式の業務以外では、null community、Garage4Hackers、hasgeek、OWASPといった、情報セキュリティの知識普及を共通の目標とするオープンコミュニティに貢献している。

イベント情報

イベント: Security Days 2025 Fall
会場: 東京会場
日時: 2025年10月23日（木）18:30-19:30
形式: ナイトセッション（軽食付き）
通訳: 同時通訳あり
公式サイト: Security Days 2025 Fall

AI Generated Summary

AI Generated Content Disclaimer

Note: This summary is AI-generated and may contain inaccuracies, errors, or omissions. If you spot any issues, please contact the site owner for corrections. Errors or omissions are unintended.

This presentation at Security Days Tokyo argues that the traditional approach of layering more security tools, scanners, and dashboards on top of growing complexity is fundamentally insufficient. Anant Shrivastava makes the case that in the age of AI-accelerated development and opaque supply chains, organizations must adopt an Attack Surface Reduction (ASR) philosophy — building systems that are simpler, smaller, and inherently less attackable through subtraction rather than accumulation. The talk covers AI’s dual impact on security, the breadth of modern supply chain risks, and a structured manifesto for reducing attack surface across software, identity, infrastructure, and data.

Key Topics Covered

The Modern Developer Workflow — Then and Now:

Previously: developers used Chrome extensions for AI prompts, VS Code with AI code-completion extensions, committed to GitHub, ran GitHub Actions, containerized to Docker, deployed on Kubernetes/EC2
Now: developers use autonomous AI agents with full command-line access, compressing the entire workflow into a single prompt — increasing speed but also risk
Shadow IT scenario: non-technical staff (Biz Dev, HR, Finance) downloading AI IDEs like Cursor/Windsurf, building and deploying applications on personal Vercel/Railway accounts without IT oversight

Defining the Attack Surface:

The total sum of all reachable ways an adversary can interact with a system
Encompasses: code (endpoints, APIs, binaries), configurations (network ports, IAM roles, CI/CD runners), data (logs, backups, analytics feeds), identities (users, tokens, service accounts), and dependencies/integrations (open-source, SaaS, AI models, APIs)

The Failure of Current Security Approaches:

Patch after compromise, add another scanner, deploy more layers, buy another dashboard, monitor more alerts — all reactive approaches that add complexity
Reality: AI has lowered the barrier for attackers, supply chains are opaque and unmanageable, cloud and CI/CD systems control everything, and misconfigurations at abstraction layers bypass traditional defenses

AI’s Impact on Security:

AI is a multiplier: it amplifies both good and bad development practices at scale
Speed negates efficiency: faster development leads to more feature requests, less testing time, more bugs, more AI-driven bug fixes — creating a self-reinforcing cycle
AI hallucinating package names creates novel supply chain attack vectors (fake packages matching hallucinated names)
AI also shows promise for defense: DARPA’s AIxCC competition demonstrated AI’s capability in finding and fixing vulnerabilities

Software Supply Chain Reality:

Organizations are approximately 80% dependent on third-party software for their needs
The supply chain problem is structural and cannot be patched with additional monitoring alone

The Attack Surface Reduction Manifesto:

Core philosophy: “Build less. Expose less. Trust less.”
Systems should be simpler, smaller, and inherently less attackable
Reduction over accumulation: rather than chasing visibility across growing complexity, advocate for inherently less complex systems
Reference: reducetheattacksurface.com

Four Areas of Attack Surface Reduction:

Software Reduction: minimize features, APIs, and endpoints; prioritize internal reuse over external dependence; shrink build and toolchain; limit codebase blast radius; harden through simplicity
IAM Reduction: principle of least privilege; human account hygiene; service identity discipline; shadow access discovery; role explosion control; federated identity risk management; secret lifecycle ASR; SaaS and vendor identity exposure management
Infrastructure Reduction: expose nothing by default; reduce control plane and runtime complexity; minimalism at image and execution level; prune zombie infrastructure and orphaned services; reduce lateral movement paths; runtime cleanup and reboot culture; infrastructure-as-code and deployment discipline
Data Reduction: collect less data by default; reduce data retention; limit data propagation and transformation; avoid over-engineering analytics

Convincing Management:

Quantifiable outcomes: reduced assets mean reduced costs for ongoing operations, compliance, electricity, data traffic, per-device licensing, and storage
Frame attack surface reduction as a cost-saving measure, not just a security initiative

Actionable Takeaways

Adopt the Attack Surface Reduction mindset: “Build less. Expose less. Trust less.” — every system, service, identity, and data store that doesn’t need to exist is one less thing to defend.
Audit and reduce your software footprint: eliminate unused features, APIs, and endpoints; prefer internal code reuse over adding external dependencies; minimize build toolchain complexity.
Address IAM sprawl systematically: enforce least privilege, discover shadow access, control role explosion, and manage the secret lifecycle proactively.
Prune infrastructure: decommission zombie resources and orphaned services, expose nothing by default, adopt minimal container images, and enforce infrastructure-as-code discipline.
Collect and retain less data: every piece of data you store expands your attack surface and compliance burden — minimize collection, shorten retention, and limit propagation.
Account for Shadow AI: establish policies for AI IDE usage, monitor for unsanctioned deployments by non-technical staff, and bring shadow IT into governance.
Make the business case with cost savings: translate attack surface reduction into reduced licensing, operational, compliance, and infrastructure costs to secure management buy-in.
Visit reducetheattacksurface.com and use the ASR Manifesto as a framework for systematically reducing exposure across software, identity, infrastructure, and data.