Vishing explained: the voice phishing threat enterprises can no longer ignore

Key insights

  • Vishing attacks surged 442% in H2 2024 and now account for over 60% of phishing-related incident response engagements, making voice phishing the dominant social engineering vector facing enterprises.
  • AI voice cloning needs just three seconds of audio to produce a convincing replica, and the FBI has issued a formal warning about AI-generated voice messages impersonating senior U.S. officials.
  • The 2025-2026 ShinyHunters/Scattered Spider campaign compromised 760+ organizations through vishing, proving that voice phishing is an enterprise-grade initial access vector, not a consumer scam.
  • Enterprise detection requires correlating voice call activity with authentication events — VoIP/SIP logs, MFA push anomalies, and remote access tool installation patterns all provide SOC-actionable signals.
  • Vishing maps to specific MITRE ATT&CK techniques (T1566.004 and T1598.004) and compliance framework controls that organizations should include in their security programs.

Vishing is accelerating faster than any other social engineering vector. According to the CrowdStrike 2025 Global Threat Report, voice phishing attacks surged 442% between the first and second halves of 2024 — and the first half of 2025 already exceeded all of 2024. Meanwhile, Cisco Talos reported that vishing accounted for over 60% of all phishing-related incident response engagements in Q1 2025, making it the most common phishing type their team encountered. For security teams still treating vishing as a consumer nuisance, the data tells a different story. Enterprise-grade campaigns — powered by AI voice cloning, social engineering playbooks, and coordinated threat actor alliances — now target SSO credentials, CRM platforms, and executive identities at scale.

What is vishing?

Vishing is a form of phishing that uses voice communication — phone calls, VoIP, or voice messages — to manipulate targets into revealing sensitive information, installing remote access tools, or transferring funds. The term combines "voice" and "phishing," placing it alongside smishing (SMS phishing) and spear phishing (targeted email phishing) in the broader social engineering taxonomy.

What makes vishing uniquely dangerous is the voice channel itself. A live phone call conveys urgency and authority more convincingly than text. Attackers exploit this by impersonating trusted figures — IT help desk staff, bank representatives, government officials, or even executives whose voices have been cloned using AI. The result is a social engineering attack that bypasses email filters, link scanners, and many of the text-based controls organizations rely on.

The scale of the problem is significant. The CrowdStrike 2025 Global Threat Report documented a 442% increase in vishing attacks between H1 and H2 2024. Cisco Talos confirmed the trend from the incident response side, finding that vishing was the most common phishing type in Q1 2025, accounting for over 60% of all phishing-related engagements. Seventy percent of organizations have fallen victim to a voice phishing attack, according to Keepnet Labs.

Vishing vs phishing vs smishing

Understanding how vishing fits within the phishing family helps security teams map the right controls to the right vectors.

How phishing, vishing, and smishing compare across delivery channel, typical lures, and defensive requirements:

Attack type Channel Common lure Key difference
Phishing Email Malicious links, fake login pages, invoice fraud Blocked by email gateways, URL filtering
Vishing Phone call / VoIP / voice message IT support impersonation, bank verification, government threats Bypasses text-based security controls; exploits voice authority
Smishing SMS / text message Package delivery, MFA codes, account alerts Exploits mobile trust; limited screen real estate hides red flags

The critical distinction for defenders: vishing operates outside the channels most security tools monitor. Email and SMS phishing leave digital artifacts — URLs, sender headers, message metadata — that detection systems can analyze. Vishing leaves a phone call and whatever the human on the other end decides to do next.

How vishing works

A typical vishing attack follows a structured chain that combines technical preparation with psychological manipulation.

  1. Reconnaissance — Attackers gather target information from LinkedIn, company directories, breached databases, and social media. This reconnaissance phase builds the pretext.
  2. Pretext development — The attacker crafts a believable scenario: an IT emergency, a bank fraud alert, a compliance audit, or an executive request.
  3. Caller ID spoofing — Using VoIP infrastructure, attackers perform caller ID spoofing to display a trusted number — the company's help desk, a known bank, or a government agency.
  4. Initial call and rapport building — The attacker establishes credibility by referencing specific details (employee name, department, recent ticket numbers) gathered during reconnaissance.
  5. Psychological manipulation — Urgency ("Your account will be locked in 15 minutes"), authority ("This is the security team"), and fear ("We've detected unauthorized access") drive the target toward compliance.
  6. Credential harvesting or tool installation — The target is directed to a credential-harvesting site, asked to share MFA codes, or instructed to install remote access tools like Quick Assist or AnyDesk.
  7. Post-compromise activity — Attackers use harvested credentials or remote access for lateral movement, data exfiltration, MFA enrollment on attacker-controlled devices, or ransomware deployment.

Vishing attackers typically seek credentials (SSO passwords, MFA codes), remote system access (through tool installation), financial information (bank details, wire transfer authorizations), and personal data that supports further social engineering.

AI-powered vishing and deepfake voice cloning

AI has transformed vishing from a manual, low-scale operation into an industrial threat. Voice cloning technology now needs just three seconds of audio to produce a convincing replica of someone's voice, based on Microsoft's VALL-E research cited in voice phishing statistics compiled by Programs.com. According to Fortune's 2026 deepfake outlook, AI-generated voice has crossed the "indistinguishable threshold" — meaning the average listener cannot reliably distinguish a cloned voice from a real one.

The implications are severe. In May 2025, the FBI issued PSA250515 warning that malicious actors were using AI-generated voice messages to impersonate senior U.S. government officials, targeting current and former federal and state officials for credential harvesting and account takeover. Deepfake-enabled fraud losses are projected to reach $40 billion by 2027, according to Deloitte estimates cited in Programs.com. This is the AI phishing evolution that security teams must prepare for.

Callback phishing and TOAD

Callback phishing — also known as telephone-oriented attack delivery (TOAD) — represents a hybrid attack chain that combines email and voice channels. The pattern, documented extensively by Cisco Talos in their Q1 2025 IR trends report, works like this: attackers first flood the target's inbox with spam or send a convincing notification email (fake subscription confirmation, invoice, or security alert). The email includes a phone number for "support." When the target calls, a live attacker walks them through installing remote access software like Quick Assist, granting the attacker direct system access.

This technique is particularly dangerous because the victim initiates the call, which feels safer than an incoming call from an unknown number. BazarCall campaigns pioneered this approach, and it has become a dominant pattern in enterprise vishing operations.

Types of vishing attacks

Vishing attacks range from mass automated campaigns to highly targeted operations. Each type presents distinct challenges for detection and prevention.

Vishing attack types ranked by sophistication and enterprise risk:

Attack type Typical target Key technique Detection difficulty
VoIP/wardialing Mass consumer Automated IVR systems dial thousands of numbers Low — high volume, generic scripts
Government impersonation Individuals, small businesses IRS, Social Security, law enforcement pretext Low-Medium — recognizable patterns
Financial institution vishing Bank customers, finance teams Account verification, fraud alert pretext Medium — uses real account details
Technical support vishing Employees, help desk staff IT impersonation, remote tool installation Medium-High — Scattered Spider procedure per MITRE ATT&CK T1566.004
Callback phishing / TOAD Enterprise employees Email spam flood followed by phone-based social engineering High — victim initiates contact
CEO/executive impersonation Finance, HR, executive assistants Deepfake voice, wire transfer or data requests High — IBM reports a $25M loss in one Hong Kong case
AI deepfake vishing Government officials, executives Real-time AI voice cloning from seconds of audio Very High — indistinguishable from authentic

Organizations with 70% reporting victimization according to Keepnet Labs face an estimated $14 million in average annual costs from vishing attacks, though this vendor-sourced figure should be interpreted cautiously given unclear methodology. The key insight is that vishing is no longer a consumer problem — it is a systematic enterprise social engineering vector that threat actors treat as a professional service.

Vishing in practice: 2025-2026 case studies

The 2025-2026 period has produced the most significant enterprise vishing campaigns on record. These examples demonstrate that voice phishing is now a primary initial access vector for advanced threat actors.

ShinyHunters/Scattered Spider campaign (2025-2026). The most impactful vishing campaign of this period targeted over 760 companies, according to ReliaQuest and Computer Weekly. The ShinyHunters collective, with initial access provided by Scattered Spider operators, used custom vishing kits to target SSO environments (Google, Microsoft, Okta). Confirmed victims include Google, Cisco, Wynn Resorts (800,000+ employee records), CarGurus (12.5 million records), and Harvard University. Picus Security characterized the alliance as one of the most dangerous cybercrime supergroups of 2025. The campaign demonstrated that vishing is now delivered as a professional service, with operators recruited at $500-$1,000 per call using pre-written scripts targeting IT help desks.

Cisco CRM breach (July 2025). A single Cisco employee was socially engineered via a vishing call, resulting in the attacker accessing and exporting profile information from a third-party cloud-based CRM system. Cisco's own security advisory confirmed the data breach, underscoring that even security-conscious organizations are vulnerable when one employee is compromised.

FBI IC3 warning on AI voice impersonation (May 2025). The FBI issued PSA250515 after discovering that malicious actors had been using AI-generated voice messages and texts to impersonate senior U.S. government officials since at least April 2025. The campaign targeted current and former federal and state officials for credential harvesting. According to Google Cloud/Mandiant's tracking, these techniques continue to evolve and expand in scope.

Harvard University breach (November 2025). Harvard's Alumni Affairs and Development systems were compromised through a vishing attack. The breach exposed alumni data and development relationships in what analysts estimated could carry significant long-term impact for the institution.

These cases share a common thread: vishing served as the initial access vector that enabled downstream compromise — credential theft, data exfiltration, and in several cases, ransom demands.

Detecting and preventing vishing

Effective vishing defense goes beyond telling employees not to answer unknown calls. It requires correlating voice channel activity with authentication events and building SOC detection capabilities that identify post-vishing compromise behavior.

Enterprise detection strategies

Because the vishing call itself occurs outside most security monitoring, SOC teams must focus on detecting the behavioral patterns that follow a successful vishing attack.

Enterprise SOC detection signals for post-vishing compromise activity:

Detection signal Data source SOC action
Remote access tool installation (Quick Assist, AnyDesk, TeamViewer) EDR / endpoint telemetry Alert and investigate; block unauthorized RAT installations
MFA push within minutes of inbound phone call Identity provider logs + VoIP/SIP logs Correlate timing; require callback verification before approval
Credential reset followed by new device MFA enrollment Identity provider / Azure AD / Okta logs Investigate identity change chain; require manager approval
Unusual data export from CRM/SaaS after identity event CASB / SaaS audit logs Correlate with identity anomalies; trigger DLP review
OAuth device code authentication from unexpected location Azure AD sign-in logs Monitor device code flow abuse — a novel vishing technique
Anomalous VPN or remote access after hours Behavioral analytics / NDR Correlate with phone-initiated identity events

This detection approach aligns with MITRE ATT&CK T1566.004 guidance, which recommends monitoring corporate device call logs for unusual numbers and correlating MFA push attempts with voice call activity.

Prevention best practices

  1. Implement callback verification procedures. Require all sensitive requests received by phone to be verified by calling back on a pre-registered, independently verified number. This single control breaks the attacker's control of the communication channel.
  2. Restrict remote access tool installation. The Cisco Talos Q1 2025 attack pattern consistently involved tricking users into installing Quick Assist. Restrict which remote access tools can be installed and by whom through application allowlisting.
  3. Enforce multi-factor authentication with phishing-resistant methods. Hardware security keys (FIDO2) resist MFA social engineering. Be aware that push-based MFA can itself be bypassed through vishing — attackers talk targets through approving the push.
  4. Deploy STIR/SHAKEN caller authentication. This telecom-level protocol verifies caller identity, helping detect spoofed numbers before they reach employees.
  5. Run vishing simulations as part of security awareness training. Organizations running regular vishing simulations achieve up to 90% attack recognition success rates, according to Keepnet Labs. However, 33% of trained employees still disclose information despite strong warnings, demonstrating that training alone does not eliminate the risk. A baseline of 6.5% of employees disclose sensitive information during simulated vishing tests.
  6. Control SSO/MFA enrollment processes. Prevent attacker device registration after credential compromise by requiring additional verification for new device enrollment.

What to do if a vishing attack succeeds

If a vishing attack compromises credentials or grants remote access, security teams should immediately rotate affected credentials, revoke active sessions, audit the compromised account for data access and changes, check for new MFA device enrollments, scan endpoints for remote access tools, and initiate a forensic investigation to determine the scope of access. Speed matters — the window between initial compromise and data exfiltration is often measured in minutes.

Vishing and compliance

Vishing maps to specific controls across major compliance frameworks — a connection that no competitor in the top search results currently makes. GRC teams should include these mappings in their audit evidence and risk assessments.

Compliance framework controls applicable to vishing risk management:

Framework Control ID Control name Vishing relevance
MITRE ATT&CK T1566.004 Spearphishing Voice (Initial Access) Direct mapping; procedure examples from Scattered Spider, Storm-1811
MITRE ATT&CK T1598.004 Spearphishing Voice (Reconnaissance) Reconnaissance-phase vishing; LAPSUS$ help desk calls
NIST CSF 2.0 PR.AT Awareness and Training Security awareness training must include vishing recognition
NIST SP 800-53 AT-2, AT-3 Awareness Training, Role-Based Training Help desk and IT support staff are primary vishing targets
CIS Controls v8 14.2, 14.5 Recognize Social Engineering, Simulated Tests Vishing simulations satisfy Control 14.5
ISO 27001:2022 A.6.3, A.5.14 Security Awareness, Information Transfer Policies on information disclosure over phone calls
PCI DSS v4.0 12.6 Security Awareness Training Training must cover threats to cardholder data including vishing

Compliance frameworks provide the structure, but they only work when organizations actually map specific threats like vishing to specific controls. CISA's social engineering guidance provides additional context for organizations building their vishing defense programs.

Modern approaches to vishing defense

The security industry is responding to the vishing surge with solutions that span prevention, detection, and response. Key categories include network detection and response (NDR) for post-vishing behavioral detection, identity threat detection and response (ITDR) for credential abuse monitoring, security awareness training platforms with vishing simulation capabilities, and emerging voice deepfake detection tools.

On the deepfake detection front, isVerified emerged from stealth in January 2026 with applications designed to identify AI-generated voice in real time. The market need is clear — more than $200 million was lost to deepfake scams in Q1 2025 alone. SecurityWeek's Cyber Insights 2026 report predicts that social engineering will evolve toward "relationship operations" — sustained, AI-assisted psychological manipulation campaigns that combine voice, text, and video channels over weeks or months rather than single calls.

How Vectra AI thinks about vishing defense

Vishing is an initial access vector. The phone call itself is difficult to prevent — but what happens after a successful vishing attack produces detectable behavioral patterns. Attack Signal Intelligence focuses on identifying the post-compromise behaviors that follow a successful vishing call: remote access tool installation, anomalous identity usage, unusual data access patterns, and lateral movement across the network. This assume-compromise philosophy means defenders gain coverage even when the voice-based social engineering succeeds, because attackers still need to act within the network to achieve their objectives — and those actions generate signal.

Future trends and emerging considerations

The vishing landscape will continue evolving rapidly over the next 12-24 months. Several developments deserve attention from security leaders.

Real-time deepfake voice during live calls. Current attacks often use pre-recorded AI-generated messages, but the technology for real-time voice conversion during live conversations is maturing. DEF CON's AI vishing competition has already demonstrated that AI can successfully social engineer targets in controlled environments, as IBM documented. As this capability becomes more accessible, the distinction between a "real" and "synthetic" caller will disappear entirely, increasing demand for behavioral detection over voice-based authentication.

Professionalization of vishing-as-a-service. The ShinyHunters/Scattered Spider alliance has demonstrated a service model where vishing operators are recruited, paid per call ($500-$1,000), and provided with pre-written scripts and targeting data. This professionalization lowers the skill barrier and increases volume. Expect vishing to follow the ransomware-as-a-service model, with specialized operators handling different stages of the attack chain.

Regulatory acceleration. New York's Department of Financial Services issued a vishing-specific advisory in February 2026 — the first state-level financial regulator to do so. As high-profile breaches accumulate, more regulators will likely mandate specific vishing controls, vishing simulation testing, and incident reporting requirements.

Device code authentication abuse. A novel technique combining vishing with OAuth 2.0 device authorization flow abuse emerged in late 2025, targeting Microsoft Entra environments. This technique bypasses traditional MFA by exploiting a legitimate authentication mechanism, and organizations should monitor for unexpected device code flow activity as a priority.

Organizations should invest in three areas: identity threat detection capabilities that correlate voice-channel activity with authentication events, phishing-resistant MFA (FIDO2 hardware keys), and regular vishing simulation programs that measure and improve employee resilience over time.

Conclusion

Vishing has evolved from a low-tech phone scam into one of the most effective initial access vectors in the enterprise threat landscape. The 442% surge in 2024, the dominance of voice phishing in incident response engagements, and the ShinyHunters/Scattered Spider campaign targeting 760+ organizations all point to the same conclusion: voice phishing demands the same level of defensive investment that email phishing has received for the past two decades.

The path forward combines three elements. First, process controls like callback verification that remove the attacker's control over the communication channel. Second, technical detection that correlates voice-channel activity with authentication events and post-compromise behavior. Third, security awareness programs that include regular vishing simulations while acknowledging that training alone cannot eliminate the risk.

For organizations ready to strengthen their vishing defense posture, explore how Vectra AI's platform detects the post-compromise behaviors that follow successful social engineering attacks — because assuming compromise is the first step toward resilience.

Related cybersecurity fundamentals

FAQs

What is vishing in simple words?

How does a vishing attack work?

What are the signs of a vishing attack?

What role does AI play in vishing attacks?

How do businesses prevent vishing attacks?

What is a callback phishing attack?

What is an example of a vishing attack?

How to report a vishing attack?

What industries are most targeted by vishing?

Can AI clone your voice for vishing?