Search: “identity-verification”

31 stories

DomainAll Cyber AIWindow24h 3d 7d 30d AllIn the wild Important only Show duplicates

CategoryAll Data breach 3 Phishing & fraud 3 Ransomware 2 Exploit / PoC 1 Malware 1

SeverityAll critical 1 high 4 medium 3 low 2 info 0

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

88 ID Verification Breaches Show the Cost of Collecting Identity Data

A report catalogs 88 ID-verification breaches since 2011 exposing at least 2.15 billion records, with 41 incidents leaking irreplaceable biometric data and documents.

A Mysterium VPN report compiles 88 documented breaches since 2011 involving identity and age-verification data, with confirmed exposure of 2.15 billion records and claimed totals of 4.54 billion. In 41 of 88 incidents, ID scans, verification selfies, fingerprints, and biometric templates leaked, data that cannot be changed after exposure. Notable cases include the Tea app's exposed selfies, Discord's ~70,000 government IDs, vendor failures at AU10TIX, Sumsub, and Persona, and national registry breaches in Argentina (45 million records) and France (11.7 million people).

Security AffairsAug 26, 08:30 UTC · 20d agoData breach

ID verification giant IDScan confirms data breach with more than 150 million driver’s licenses stolen

ID verification firm IDScan confirms hackers stole over 150 million driver's license records, including names and license numbers, from its cloud.

IDScan confirmed that threat actors stole driver's license records — full names, license numbers, and other government ID numbers — from the company's cloud systems during a roughly year-long intrusion. Brian Krebs previously reported a dark web service exposing data on over 150 million US and Canada residents, including license photos and records of high-profile individuals such as Defense Secretary Pete Hegseth. The FBI is investigating, and IDScan indicated hackers apparently demanded payment for full access to the stolen cache; the investigation is ongoing.

TechCrunch · Securityupdated Sep 11, 09:59 UTC · 4d agofirst Sep 10, 13:21 UTC · 5d agoData breach in the wild 4 sources1

ShinyHunters claims Florida DMV breach, puts data on the clock

ShinyHunters claims it breached Florida DMV's DAVID database, stole 200,000+ driver records including SSNs, and set a September 11 extortion deadline.

The ShinyHunters extortion group claims it breached the Florida Department of Highway Safety and Motor Vehicles' DAVID driver and vehicle database and stole more than 200,000 records. As evidence it published a screenshot of a Jeffrey Epstein record showing address, Social Security number, date of birth, license number and registered vehicles, and set a September 11 deadline before publication. The group says it obtained access through a password-reset weakness, compromised employee accounts, and queried and downloaded driver records and images. The Florida DMV has not confirmed the claim; it follows a separate confirmed IDScan.net breach exposing over 153 million license scans that prompted an FBI investigation.

CSO Onlineupdated Sep 11, 20:00 UTC · 4d agofirst Sep 9, 11:25 UTC · 6d agoData breach in the wild 3 sources1

It sure looks like hackers breached a major ID card verification service

Suspected breach of ID verification firm IDScan exposed over 150 million US and Canadian driver's licenses and passports, sold via dark web site Nexus.

Independent journalist Brian Krebs reported that a dark web site called Nexus is selling searchable access to more than 150 million driver's licenses and passports belonging to people in the United States and Canada, adding roughly 500,000 documents daily from a 'major identity verification company.' Krebs and researcher Zach Edwards identified Louisiana-based identity verification service IDScan as the likely source; Krebs' own license and Defense Secretary Pete Hegseth's records appeared in the database. The FBI confirmed it is looking into the incident and IDScan says it is investigating; Nexus went offline shortly after the report went live.

TechCrunch · SecuritySep 2, 19:35 UTC · 13d agoData breach in the wild

Revolut Data Breach Exposes Customers’ Passport Copies and Full Transaction Histories to Hackers

Revolut leaked KYC documents and full transaction histories after a fraudulent, domain-authenticated email request impersonating a government agency.

Revolut disclosed that an attacker using an unauthorized email account on a legitimate government domain, with valid domain-authentication credentials, tricked the fintech into releasing customer data. The exposed data includes passport and driver's license copies, identity-verification selfies, full names, dates of birth, addresses, IBANs, and complete transaction histories including Bitcoin activity. Revolut says core systems, accounts, and funds were not compromised, and it blocked the email source and notified authorities. On-chain investigator ZachXBT and others indicated the operation targeted high-net-worth users facing elevated phishing, SIM-swap, and extortion risk.

Cyber Security NewsSep 13, 05:09 UTC · 2d agoData breach3· 1 read

Key Reasons Why Identity Fabric Matters in 2026

Identity sprawl and unowned machine identities leave enterprise access unobserved at runtime; identity fabrics aim to close the gap between policy intent and execution.

This sponsored explainer describes identity fabric as an architectural approach connecting identity providers, governance systems, applications, and infrastructure into one observable layer that compares designed access intent with runtime execution. It argues identity sprawl across SaaS, APIs, and cloud workloads, plus unmanaged non-human identities (service accounts, bots, workloads, API keys), leaves overprivileged, dormant, and unowned machine identities unmonitored. IdP-only monitoring misses application-layer attacks, and the piece advocates behavioral visibility and lifecycle governance for secrets and machine identities.

The Hacker NewsAug 28, 11:30 UTC · 18d agoIndustry

NovaCookies Campaigns Abuse Genuine Docusign Notifications to Steal Microsoft 365 Sessions

NovaCookies AitM phishing kit, a Sneaky 2FA variant, uses genuine Docusign lures to steal Microsoft 365 sessions at hundreds of organizations.

Island disclosed NovaCookies, a $320/month adversary-in-the-middle phishing-as-a-service platform that relays Microsoft 365 sign-ins through attacker infrastructure to capture credentials, MFA codes, and authenticated sessions. Campaigns abuse genuine Docusign envelopes and Microsoft/Google redirect hops so each step looks legitimate, with lure domains on .vu and alternating-case labels such as PwPt-sHaRe. Proofpoint assesses NovaCookies as a Sneaky 2FA variant with added flows for Okta and Entra domains federated to GoDaddy, and a fully managed PhaaS model. It has targeted hundreds of organizations in the U.S., U.K., Canada, Germany, Israel, and the U.A.E., and is advertised via Telegram with anti-analysis checks like a Cloudflare gate.

The Hacker NewsSep 1, 13:02 UTC · 14d agoPhishing & fraud

Access Control as Verified Parse Constraints

Researchers verify a class of EverParse validators that correctly enforce access-control policies, deploying a machine-checked enforcement gate on seL4.

The paper targets enforcement-code bugs in commercial security gateways by proving that forward-only, backtrack-free EverParse validators are verified recognizers for a bounded finite-state class that includes access-control decision functions with fixed-offset fields and bounded disjunction. Encoding a bounded policy language into a fixed-size byte buffer allows an SMT solver to verify the enforcement code once, covering all byte values, policies, requests, and sessions. Editing rule content over a fixed endpoint set requires no new proof, while adding endpoints reruns the toolchain. A deployment on the seL4 microkernel ensures every request passes through the gate and unverified components cannot corrupt the enforcement chain.

arXiv cs.CRSep 11, 06:33 UTC · 4d agoResearch

Weekly Cybersecurity Newsletter Bulletin – Microsoft 0-day, FortiOS, PAN-OS Flaw, Revolut Data Breach, and 20+ Stories

Weekly roundup: Microsoft patches 973 flaws including two actively exploited zero-days; FortiOS CAPWAP flaw deploys PivotC2 RAT; PAN-OS root RCE disclosed.

Microsoft's September 2026 Patch Tuesday fixed 973 vulnerabilities, including two zero-days under active exploitation: CVE-2026-85880 (Windows ALPC) and CVE-2026-81963 (Windows Update Stack), both elevation-of-privilege bugs. SOCRadar reported active exploitation of CVE-2025-25249 (CVSS 9.8) in FortiOS CAPWAP, deploying a Node.js RAT called PivotC2 that exfiltrates Exchange mailboxes to Wasabi cloud storage; 178 devices were compromised out of 30,000 scanned IPs, attributed to a Russian-speaking financially motivated group. Palo Alto disclosed CVE-2026-0310, a 9.2-rated buffer overflow enabling root code execution on PA-Series firewalls, and Fortinet disclosed CVE-2026-84393, a ZTNA certificate validation MITM flaw. Cyera also revealed CVE-2026-6471 ('PostGREShell'), a 12-year-old PostgreSQL logical-decoding flaw allowing code execution via REPLICATION-privileged accounts.

Cyber Security NewsSep 14, 03:10 UTC · 1d agoExploit / PoC in the wildCVE-2026-85880 CVE-2026-81963 CVE-2025-25249+3 CVEs2

How to Spot and Stop Rogue Device Joins

Wiz details how adversaries use realistic device names in Entra ID registrations to evade detection and which behavioral signals still expose them.

Wiz researchers describe how adversaries generate realistic device names during Entra ID device registration instead of leaving recognizable fingerprints from public tooling, making rogue device joins blend into enterprise environments. The post explains how this trend changes Entra ID detection approaches. It also identifies behavioral signals that still expose these attacks. No specific CVE or victim was named.

Wiz BlogAug 18, 16:24 UTC · 28d agoResearch in the wild

AI is making fraud harder to spot and identity harder to prove

Experian's 2026 report finds AI-generated scams, deepfakes and synthetic identities spreading across digital channels, pushing businesses toward adaptive identity verification and 'Know Your Agent' checks.

Experian's 2026 US Identity & Fraud Report, based on consumer and business surveys, found 60% of consumers aware of AI-generated image/video scams, 53% of AI phishing, and 47% of deepfake voice impersonation. 80% of US businesses already use machine learning or generative AI in fraud management, while AI chatbot account openings rose from 16% in 2025 to 27%. The report highlights adaptive authentication, behavioral biometrics, and emerging 'Know Your Agent' controls as AI agents begin acting on behalf of customers.

Help Net SecurityAug 20, 00:00 UTC · 26d agoPhishing & fraud

Subgroup Membership Inference Audits of Differentially Private Synthetic Text

Audits of 32 differentially private synthetic-text releases show subgroup membership leakage is concentrated in few records and systematically underestimated by average-case attacks.

The paper defines a subgroup-targeted membership inference game in which the target pool is an explicit parameter, to audit residual leakage in differentially private synthetic text releases. The audit instantiates 32 proxies across four datasets, three generators (DP-SGD fine-tuning, API-based prompting, and activation steering), and five privacy budgets. DP substantially reduces average leakage at every budget, but remaining leakage is concentrated: roughly a tenth of records carries about 40% of it, and the noise removes more measured leakage from random records than from high-risk ones. Which records leak depends on the release mechanism, so record-level risk cannot be assessed independently of the release.

arXiv cs.CRSep 9, 07:56 UTC · 6d agoResearch

Predicting Privacy Leakage from Weight Spectral Density

Study shows WeightWatcher spectral metrics like stable rank correlate with membership inference vulnerability, enabling cheaper ML privacy auditing.

The paper tests whether spectral metrics from the heavy-tailed self-regularisation framework can proxy membership inference attack (MIA) vulnerability without training expensive shadow models. On image and tabular classification tasks, stable rank correlates positively with overall MIA success, while Log alpha-Norm correlates negatively at the low false-positive regime. These correlations are stronger than those obtained from the generalisation gap, suggesting weight spectra capture leakage information overfitting measures miss. The authors propose spectral analysis as a scalable direction for privacy auditing.

arXiv cs.CRSep 10, 16:34 UTC · 5d agoResearch

Topological Fraud Detection in Latent Transaction Spaces

Researchers present a privacy-preserving fraud detection method combining unsupervised filtering and supervised classification on anonymized transaction embeddings for low-latency triage.

The paper describes fraud detection performed entirely on topologically anonymized transaction embeddings. It iterates unsupervised filtering followed by supervised classification ('sniping') to flag suspicious activity. The goal is ultra-low-latency, privacy-preserving triage for institutions without exposing personally identifiable information.

arXiv cs.CRSep 8, 08:50 UTC · 7d agoResearch

The Machine With Many Faces: Post-Exploitation Identity Misuse in SPIFFE/SPIRE

Unit 42 demonstrates that root access on a Kubernetes node lets attackers spoof SPIFFE/SPIRE attestation and harvest co-located workloads' SVIDs.

Palo Alto Networks Unit 42 describes post-exploitation techniques in which an attacker with root on a compromised Kubernetes node spoofs Linux cgroup metadata used by the SPIRE agent during workload attestation, tricking it into issuing a co-located workload's SPIFFE Verifiable Identity Document to an attacker-controlled process. The research shows the core trust assumption of machine-identity systems—that the node is trusted—collapses once root is obtained, exposing all cryptographic identities scoped to that node. Unit 42 released an open-source tool, Spooffe, for defenders to test identity exposure, and notes the technique has not been observed exploited in the wild.

Palo Alto Unit 42updated Sep 11, 07:24 UTC · 4d agofirst Sep 10, 10:00 UTC · 5d agoResearch 10 sources1

When the Attacker Wears Your Logo: Detecting and Taking Down Impersonation at AI Speed

Cyble details how attackers use lookalike domains and cloned sites for brand impersonation scams and how AI-speed detection enables rapid takedowns.

Cyble describes the growing risk of digital impersonation, where attackers register lookalike domains, clone websites, create fake executive profiles, publish fraudulent job ads, and imitate customer-support accounts without ever breaching the legitimate organization. The goal is to borrow a trusted brand's credibility to make scams look legitimate, with particular risk for professional services, financial, legal, and consulting firms. The piece outlines AI-assisted approaches for detecting and taking down impersonation infrastructure quickly.

CybleAug 19, 14:01 UTC · 27d agoPhishing & fraud

Ping Identity introduces enterprise security for personal AI agents

Ping Identity launched Enterprise Personal Agent Access to discover, authenticate and control personal AI agents like Claude at runtime.

Ping Identity announced Enterprise Personal Agent Access, delivered through PingOne Privilege, to give enterprises visibility and control over personal AI agents such as Claude and Claude Code. The product discovers agents including shadow AI, associates each session with the user and device behind it, and enforces allow, deny, log, human-approval or revocation decisions on agent access at runtime. It enables secretless developer workflows so agents can commit code without long-lived credentials while keeping actions attributable to the agent. Ping cites Gravitee research that 48% of production AI agents run unsecured and says the product is available now and piloting with global enterprises.

Help Net SecuritySep 3, 06:25 UTC · 12d agoAI safety & security

CISA and NIST Release Technical Checklist for Safeguarding Identity Tokens From Theft and Misuse

CISA and NIST published NIST IR 8587, final guidance for protecting identity tokens from forgery, theft, replay, and signing-key compromise.

NIST Interagency Report 8587 (September 15, 2026) expands the IA-13 'Identity Providers and Authorization Servers' control from NIST SP 800-53 R5.1.1, guiding federal agencies and cloud providers on SSO, identity federation, and machine-to-machine authentication. It requires hardware-backed signing-key storage for moderate-impact systems, 90-day key rotation for high-impact systems, token lifetimes under one hour, and sender-constrained mechanisms such as mutual TLS and DPoP. The report cites incidents including forged SAML assertions that exposed over 60,000 emails from a federal agency. It also extends guidance to agentic AI systems using signed tokens and urges post-quantum cryptography migration planning.

Cyber Security News17:04 UTC · 6h agoAdvisory

You've Got a BUD in Me: Authenticated Reads from Per-Block Write Logs

Researchers propose BUD, per-block write-log digests enabling blockchain validators to serve historical membership and exclusion proofs far cheaper than state-wide tries.

The paper introduces Block Update Digests (BUD), which authenticate each block's write log with predecessor pointers, plus a SuperBUD and exponential hierarchy to turn long unchanged intervals into short proofs. Soundness against adversarial provers and up to f Byzantine validators is proven under archive, attestation, and committee evidence assumptions. Benchmarks show a 50x state-size increase raises the base-BUD path only 1.24x versus 3.1x for in-memory and 69.5x for disk-backed Merkle Patricia tries, with read payloads below 800 bytes and p99 warm verification at 146 microseconds.

arXiv cs.CRSep 10, 08:46 UTC · 5d agoResearch

Protecting Tokens and Assertions from Forgery, Theft, and Misuse: Implementation Recommendations for Agencies and Cloud Service Providers

NIST and CISA publish final interagency report with implementation guidance for protecting tokens and assertions from forgery and misuse.

CISA released a final NIST/CISA interagency report guiding federal agencies and cloud service providers on protecting identity assertions, access tokens, and cryptographic mechanisms underlying modern authentication and authorization. It addresses forgery, theft, and misuse of signed tokens that adversaries use for lateral movement and data access in hybrid and multi-cloud, SSO, federation, and API-based environments. The final version updates token validation, secrets management, and detection-at-scale guidance gathered via the Joint Cyber Defense Collaborative, and supports Executive Order 14306 and Secure by Design principles.

CISA Advisories12:00 UTC · 11h agoAdvisory

Signal's new security feature checks if your encrypted chats were tampered with

Signal adds automatic key verification, a key-transparency system audited by Cloudflare and Trail of Bits that detects key-directory tampering in E2EE chats.

Signal launched automatic key verification, a key-transparency system that confirms a phone number or username remains bound to a consistent encryption key across Signal's key directory. Cloudflare and Trail of Bits independently audit the system, with cryptographic protection preventing auditors from seeing users' identifiers in plaintext. The feature mitigates key substitution by a compromised directory or privileged insider, though it currently requires the contact's phone number and users can fall back to manual Safety Number verification.

Help Net SecurityAug 12, 00:00 UTC · Aug 12, 2026Tools

Who's governing your AI? A trust framework for enterprise agents and models

DigiCert pitches AI Trust framework using PKI, DNS policy records and workload identity to govern shadow AI agents across enterprises.

The Register-sponsored piece outlines DigiCert's AI Trust framework for governing AI agents, built on PKI, DNS, and attestation, citing IBM's 2026 Cost of a Data Breach report that 68% of organizations lack AI governance or shadow AI detection. The approach treats agent identity as workload identity aligned with IETF WIMSE, NIST CSF 2.0, and SPIFFE/SPIRE, using short-lived credentials instead of static API keys. DigiCert also proposes DMARC-style DNS agent policy records and an AI Agent Passport cryptographically binding agent identity to approved operations, with a unified kill switch.

The Register · Security15:00 UTC · 8h agoAI safety & security

Unmasking Cloud Identities: From Behavioral Clustering to Automated Detection

Unit 42 clusters behavior of 40,000+ AWS identities from 125 cloud environments to map functional roles and enable lightweight SQL-based detection.

Palo Alto Unit 42 built an unsupervised behavioral clustering model using UMAP and HDBSCAN on AWS CloudTrail logs to map cloud identities to functional roles such as administrators, backup services, security tooling and DevOps. The study analyzed over 40,000 identities across 125 cloud environments over two months. The researchers show that heuristics extracted from the clustering map can be implemented in standard SQL, enabling role classification at scale without running a continuous ML pipeline. The methodology extends to audit logs from other cloud providers, SaaS and Kubernetes.

Palo Alto Unit 42Sep 14, 10:00 UTC · 1d agoResearch

Do AI Coding Assistants Check Before They Install? A Pre-Registered Demand-Side Audit of Trust Signals in the Research Software Supply Chain

Pre-registered audit finds AI coding assistants verified provenance signals in only 9 of 1,920 trials before installing research software packages.

The study tested whether AI coding assistants check machine-readable trust signals such as SBOMs, signed releases, and provenance attestations before installing six open-source research software projects spanning HPC and quantum computing. Three models under two operating modes produced 1,920 registered trials scored from container logs. Provenance signals were opened in only 9 of 1,920 trials (0.5%) and zero of 384 control trials, with no trial running a verification command. The authors conclude publishing signals is insufficient and verification must be built into the program running the assistant.

arXiv cs.AI / cs.LG / cs.CLSep 7, 16:55 UTC · 8d agoAI safety & security

Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys

Researchers show split-LLM training leaks privacy via zero-valued gradients on decoy rows, exposing which activations are real despite passing forward-channel checks.

A systems-security case study of a two-node split-LLM training setup found that the returned output gradient from an Untrusted Cloud Node is exactly zero for decoy rows, revealing which rows are real. Across nine seeds, zero patterns identified real rows in 4,096 of 4,096 frames per run, and an attack on frame contents recovered 0.65 to 1.50 percentage points of extra tokens over a baseline. Both datasets passed forward-channel privacy and quality checks but failed once the returned gradient was included. Row-wise gradient clipping and noise closed the leak for roughly 0.01 nats of held-out cross-entropy, though five unmeasured attack classes remain.

Hugging Face daily papersSep 2, 20:00 UTC · 13d agoAI safety & security

Scaling Verification of Cryptographic Software with Aeneas, Rust, and Lean

Microsoft SymCrypt implementations of SHA-3 and ML-KEM verified in Lean via Aeneas-extracted Rust models, with AI agents writing proofs.

The paper develops a methodology for verifying production Rust cryptographic code by using Aeneas to extract pure models into Lean, avoiding low-level pointer and aliasing reasoning. Applied to Microsoft's SymCrypt, it verifies SHA-3 and ML-KEM implementations ported from C to Rust and extends SymCrypt with FrodoKEM, ML-DSA, and HPKE. A 237 KLOC Lean development establishes safety, panic-freedom, and functional correctness of 16.7 KLOC of Rust supporting post-quantum cipher suites on x86-64 and ARM. AI agents autonomously write formal proofs verified by the Lean kernel, and evaluation shows verified Rust meets SymCrypt's performance and portability requirements.

arXiv cs.CRSep 14, 14:32 UTC · 1d agoResearch

Self-Verifying Anomaly Detection using Explainable AI for Cybersecurity of DER Networks

ExCYDER framework self-verifies anomaly detection alerts for DER power grids using LightGBM and SHAP, reaching over 98% detection accuracy.

The paper presents ExCYDER, an explainable AI anomaly detection framework for Distributed Energy Resource networks that combines LightGBM with SHAP to validate whether each model decision aligns with its feature-attribution evidence. On a realistic DNP3 dataset it achieved over 98% detection accuracy, 14.5 ms SHAP latency per alert, and confidence deviation within 5%. The self-verifying mechanism distinguishes coherent from inconsistent alerts, improving interpretability and auditability for DER-focused security operations centers.

arXiv cs.CRSep 11, 00:19 UTC · 4d agoResearch

Differential Privacy Meets Fixed Parameter Tractability: Algorithms and Lower Bounds

Theory paper combines differential privacy with fixed-parameter tractable encoders, improving approximation guarantees for combinatorial optimization and proving new lower bounds.

The paper studies combinatorial optimization under epsilon-differential privacy within the implicit encoder-decoder framework of Gupta et al. (SODA 2010), generalizing it to allow fixed-parameter tractable encoders. This circumvents approximation barriers inherent to polynomial-time algorithms and yields improved guarantees for fundamental combinatorial optimization problems. The authors establish the first representation-independent lower bounds: assuming a non-uniform variant of the Gap Exponential Time Hypothesis, no epsilon-DP encoder-decoder pair can achieve certain approximation guarantees with a subexponential-time decoder for sufficiently small epsilon. Representation-dependent lower bounds are also provided for larger epsilon.

arXiv cs.CRSep 11, 07:13 UTC · 4d agoResearch

Sound Debloating of Redundant Checks in Zero-Knowledge Machine-Learning Circuits

Automated framework soundly removes up to 48.7% of redundant constraints in ezkl and zkml ZK-ML circuits, cutting prover time by up to 72.8%.

The framework uses whole-circuit abstract interpretation and a provenance graph to verify that each removed redundant check (range proofs, sign lookups, bit decompositions) remains entailed by the rest of the circuit, provably preserving soundness. It was evaluated on MLP, CNN, RNN, and transformer circuits generated by ezkl and zkml, with up to 25.3 million constraints. It removes up to 48.7% of constraints and reduces prover time by up to 72.8% without weakening security. Under-constrained circuits in deployed ZK systems have previously enabled attackers to forge transactions and bypass verification.

arXiv cs.CRSep 9, 13:27 UTC · 6d agoResearch1

What Good Identity Hardening Looks Like

Huntress guidance defines mature identity hardening beyond baseline MFA, covering exception cleanup and detecting identity configuration drift before attackers do.

Huntress argues MFA is only a starting point and outlines what mature identity hardening actually looks like in practice. The guidance covers closing MFA exceptions and coverage gaps that create unauthenticated attack paths. It also stresses catching identity configuration drift before attackers can exploit it.

HuntressAug 25, 13:00 UTC · 21d agoResearch

SpecGuard: Inference-Time Backdoor Detection For Free

SpecGuard detects backdoored LLM behavior at inference time using speculative decoding acceptance rates, adding no extra model computation.

Researchers propose SpecGuard, an inference-time backdoor detector that repurposes draft-token acceptance rates from speculative decoding as a detection signal at zero added model-computation cost. When a trigger shifts the target model toward attacker-controlled behavior, the clean draft model's acceptance rate changes, exposing the backdoor; the paper formalizes when this signal appears and shows suppressing it weakens the backdoor. Experiments across diverse backdoor types and model families show reliable detection, including stealthy cases invisible to input-level filters. Speculative decoding is positioned as a free, always-on monitor for frequently updated deployed models.

arXiv cs.CRSep 10, 16:51 UTC · 5d agoAI safety & security 2 sources2