How Are Security Job Postings Screening and Authenticating Candidates? Scam Warnings Rose, and a Few Employers Now Name Identity Checks

3The numbers

4Identity checks and camera-on interviews

Two kinds of language count here. The first is a step that checks who the candidate is. General Dynamics Information Technology writes: "As part of the hiring process, we will ask you to complete an identity verification process that leverages advanced biometrics and artificial intelligence to ensure authenticity and protect against identity fraud." Allstate asks: "Please also have a Valid Photo Identification available at the start of your interview."

The second is a requirement that the candidate be seen. Figma writes: "To ensure the integrity of our hiring process and facilitate a more personal connection, we require all candidates keep their cameras on during video interviews." Cloudflare tells applicants at the offer stage that they "may be asked to attend an in-person interview within one of the Cloudflare Offices or Cloudflare Hubs."

General Dynamics Information Technology and Cloudflare carry 62 of the 115 postings, 54%. Without them, 1.2% of 2026 postings name a step, still a likely but borderline rise (+0.5 to +2.1). Counted once per employer, 2.2% of employers name one, a clear rise (+1.2 to +3.1). The 2024 figure stays at zero among the 131 postings applied to through a company job board. Their text kept the equal-opportunity statement, a common closing line, 67.2% of the time.

5Background checks and fraud screening

Background checks were already named in 2024 postings, and the rate did not measurably move: 4.6% of postings in 2024, 6.2% in 2026 (−1.4 to +4.7). OpenAI and Arctic Wolf carry 21% of the 2026 mentions. Security clearance investigations count here.

Fraud-screening language is rarer, and almost absent from the 2024 postings. Samsara names "a fraud detection tool, to validate the authenticity of applications and protect against identity fraud." Marvell bars AI tools in interviews, and Micron Technology warns that misrepresented qualifications disqualify a candidate. Together these reach 2.0% of 2026 postings, and General Dynamics Information Technology and Micron Technology carry 53% of them.

6Recruiting-scam and impersonation warnings

These warnings sit in the closing text of a posting. Hewlett Packard Enterprise writes: "We have become aware of an increase in fraudulent recruitment activities in which individuals impersonate our company or authorized recruitment agencies to offer fake employment opportunities." Most name the only email domain recruiters use, or state that the company never asks candidates for money.

The warning appears in 11.3% of 2026 postings from 81 employers. Anthropic and Anduril Industries carry 15% of the 517.

The source matters in both years. LinkedIn postings not applied through a company job board keep the equal-opportunity statement about half as often, so the 2024 rate may undercount warnings. Set against the board-applied 2024 postings alone, all 2026 postings, which come from company job boards, still show a rise in the warning from 2.3% to 11.3% (+3.9 to +13.8). In 2026, Lever postings in this data carry almost no closing text, and their 0.0% may undercount the other way.

The same employers point the same direction. At the companies with postings in both years, the warning went from 3.6% of their 2024 postings to 13.7% of their 2026 postings, a likely but borderline rise (+1.0 to +20.7).

7Where the data comes from

The 2024 postings come from the LinkedIn Job Postings dataset (Hugging Face datastax/linkedin_job_listings [2]): 1,038 US postings from 679 companies. The 2026 postings come from AKA's daily read of company job boards: 4,587 security postings from 831 US-headquartered companies, open on or after 23 September 2026. Before 21 September the collector skipped several non-engineering titles, so postings that closed earlier are left out. Postings by job aggregators, which are not employers, are left out in both years. Both years hold the posting text each collection kept.

Both years are counted the same way. A posting counts when its title passes the series' security title rules, a local model reads the posting and judges it a security role, and it is the only posting with that title at that company.

Each measure is a search rule applied to that text, written once and used for both years. An identity step counts when identity verification, a photo identification or an identity check sits in the same sentence as the hiring process, a candidate, an interview or onboarding. The federal I-9 formula ("all persons hired will be required to verify identity and eligibility to work") does not count as an identity step. It falls under the E-Verify and I-9 line with other eligibility boilerplate. Background checks are counted apart from identity steps. Drug tests alone and work-authorization lines are not counted.

The rules were checked by reading postings. For each measure we read 20 randomly chosen matching postings. All 20 were correct for identity steps, interview requirements, fraud screening, scam warnings and eligibility lines, and 19 of 20 for background checks. The one error, a scam warning that listed background checks among fees, is excluded by the final rule. We also read 22 near misses for each measure below, postings that came close to matching but did not. For interview requirements, 1 of 22 should have counted: an in-person interview listed as a stage without requirement wording. Fraud screening missed 1 of 22. Identity steps, background checks and scam warnings missed none.

A posting cannot show the screening that actually happens, or anything after the posting: the interview itself, the check a vendor runs, or a warning sent by email. A posting that names no step may still run one. No causal claim follows from these rates. This read covers security job postings only, although the question applies to every job. A version across all jobs will follow once the collection covers every job description.

8How this answer was stress-tested

9What would strengthen this answer