Packet Path/

Target 2013: The Heating Contractor's Password That Opened the Registers

In November 2013, attackers did not break into Target — one of the largest retailers in the United States — through its front door. They used a password stolen from a small Pennsylvania heating-and-refrigeration contractor that had remote access for billing, walked from the edge of the network to the checkout registers, and installed malware that copied card numbers out of the registers' memory as customers swiped. Target confirmed about 40 million card accounts exposed, then disclosed personal information for up to 70 million more customers. Its own detection software raised urgent alerts while the theft tools were being installed — and nobody acted on them. This is the documented story, from the Senate's kill-chain report and the court settlements that followed.

Case Studies · Intermediate · 12 min · November 9, 2026

Editorial illustration on a dark teal background: a row of glowing checkout registers in a dark retail store, with one small heating vent on the wall glowing alarming red and a faint dotted line running from the vent across the wall to the registers.

The short answer

In November and December 2013, attackers stole the network credentials of Fazio Mechanical Services, a small heating, ventilation and air conditioning (HVAC) contractor in Sharpsburg, Pennsylvania, that had remote access to Target's network for electronic billing, contract submission, and project management. Using those credentials, the attackers reached Target's internal network by November 12, 2013, moved from that low-sensitivity foothold to the point-of-sale (POS) registers — the checkout terminals — and installed a tailored version of BlackPOS, malware that scrapes card data out of a register's memory (a "RAM-scraping" attack — RAM is a computer's short-term working memory) in the brief moment when the data sits there unencrypted. Target publicly confirmed on December 19, 2013 that about 40 million credit and debit card accounts were exposed, and on January 10, 2014 disclosed that names, addresses, phone numbers, and email addresses for up to 70 million customers were also stolen. Target's FireEye anti-intrusion software raised urgent alerts while the attackers' data-theft malware was being installed — and, as the U.S. Senate's report puts it, Target missed a series of opportunities to stop the attack at almost every step. The legal bill that can be counted from court and settlement documents includes a $10 million consumer settlement (final approval November 17, 2015) and an $18.5 million settlement with 47 states and the District of Columbia (May 23, 2017) that also forced a written security program, network segmentation, and two-factor authentication.

The company: what Target did

Target is one of the largest retail companies in the United States — a national chain of big stores selling everything from groceries to electronics, taking payment by credit and debit card at checkout registers in every store. Why that matters here: a retailer that size processes an enormous number of card payments every day, especially in the weeks before Christmas. Every one of those payments passes, for a fraction of a second, through the memory of a checkout register.

Two pieces of vocabulary carry this whole story. A point-of-sale (POS) terminal is the register device that reads your card and sends the payment for processing. And network segmentation means dividing a company's computer network into separated zones — like watertight compartments in a ship — so that getting into one zone does not give you a path into the others. Target's card-payment zone was supposed to be the most protected compartment of all. Target had even been certified in September 2013 — weeks before the breach — as compliant with the Payment Card Industry Data Security Standard (PCI DSS), the rulebook card companies require of merchants who process card payments. The breach happened anyway. This case is about the distance between having the tools and the certificate, and having the compartments, checks, and responses that actually stop an intruder.

One honest note about sources before the story starts, because it shapes how this page is written. The fullest single account is the U.S. Senate Commerce Committee's March 26, 2014 report, A "Kill Chain" Analysis of the 2013 Target Data Breach. The report says plainly, on its first page, that it explains the breach based on media reports and expert analyses published since Target acknowledged the breach — Target's own forensic examination was not complete, and some steps (especially exactly how the attackers travelled from the billing system to the registers) were not fully clear in the public record. Where this page repeats a step the report itself flags as a theory, it says so. The money figures, by contrast, come from court and settlement documents, which are exact.

The problem: a chain of small doors, all unlocked

The Senate report organises the breach using the intrusion kill chain — a framework from Lockheed Martin security researchers (2011) that breaks an attack into steps: reconnaissance, weaponisation, delivery, exploitation, installation, command and control, and finally the action itself. The point of the framework is hopeful: an attacker has to complete every step, while a defender only has to break one. The report's verdict is that Target had a chance at nearly every step. Five weaknesses lined up:

First, a vendor with weak security held a real key. Fazio Mechanical Services had remote access to Target's network for billing and project work. Investigators familiar with the case reported that Fazio used the free version of Malwarebytes Anti-Malware — a consumer tool with no real-time protection. Attackers sent Fazio malware-laden phishing emails (emails designed to trick a reader into opening an infected attachment) at least two months before the breach began, and reportedly used password-stealing malware — identified in reporting as Citadel, though the Senate report notes this was not confirmed — to steal Fazio's Target credentials.

Second, a password was the only lock. Two-factor authentication means a login needs a second proof beyond a password — for example, a code sent to a phone. According to a former Target vendor manager cited in the report, Target rarely required two-factor authentication from its low-level contractors. So one stolen password from a small contractor was, by itself, enough to log in as that contractor.

Third, the compartments leaked. The billing system the vendor credentials opened — Target's Ariba external billing system — sits at the edge of the network and holds no card data. Yet the attackers ended up on internal servers and on the registers themselves. The report concludes that the outside portal was likely not fully isolated from the rest of Target's network, and that Target failed to properly isolate its most sensitive network assets. How exactly the attackers crossed over is one of the steps the report says is not fully clear; one reported theory is that they exploited a default account (a built-in login that ships with software) in a BMC Software IT-management product. Treat that as a theory, not a finding — but note that changing or removing default accounts was already a PCI DSS requirement.

Fourth, the registers held plaintext gold for a split second. When you swipe a card, the card's data passes through the register's memory unencrypted, in plaintext, before being sent on to the payment processor. Malware that reads that memory at that moment — RAM scraping — gets everything. This was not a new trick: Visa had published warnings about RAM-scraping malware targeting retailers in April and August 2013, months before the breach.

Fifth, the alarm system reported to no one who acted. Target had layers of protection — firewalls, malware detection, intrusion detection and prevention, data-loss prevention tools — and its FireEye system did exactly its job, raising urgent alerts as the attackers installed their theft tools. But the alerts were not acted on, and FireEye was not allowed to delete the malware automatically. An alarm that nobody answers is not a control; it is a log of the moment you could have stopped it.

An analogy: imagine a warehouse where a cleaning contractor's badge opens the loading dock, the badge needs no second check, the dock door into the main floor is left ajar, the safe's contents sit briefly on an open desk every time they are counted, and the motion sensors phone a guard desk that has decided the sensors cry wolf. No single failure burgles the warehouse. All five together do.

The path in: five steps, each one stoppable 1 · CREDENTIALS STOLEN — a phishing email at the HVAC vendor Fazio Mechanical's Target password is stolen at least two months before the breach 2 · GATEWAY ACCESS — November 12, 2013 No second factor: the stolen password alone logs the attackers in as the vendor 3 · MOVEMENT INSIDE — from the billing edge to the card zone The Senate report: the sensitive assets were not properly isolated from the rest 4 · BLACKPOS ON THE REGISTERS — tested Nov 15–28, most registers by Nov 30 RAM scraping copies plaintext card data from register memory at the moment of a swipe 5 · EXFILTRATION — from December 2, by FTP, in business hours 11 GB collected via a Russia-based server; transfers ran 10:00 a.m.–6:00 p.m. Central

What this shows: the kill chain as the Senate report reconstructs it. The attackers needed all five steps. Target had a realistic chance to break the chain at every one of them — vendor security, a second login factor, segmentation, register protection, and alert response.

What happened, September 2013 – January 2014

Dates are as given in the Senate report and its timeline, except where a settlement document states its own finding. Where reports differ on a detail, this page says which source it is following.

The breach, on the Senate report's timeline September 2013 — certified PCI DSS compliant Weeks later, phishing emails steal the HVAC vendor's Target credentials November 12 — attackers first access the internal network The states' investigation: gateway server reached through the vendor credentials November 15–28 — POS malware tested on a few terminals Symantec also detects malicious behaviour around November 28 — on the same server November 30 — most registers infected; theft tools installed FireEye raises urgent alerts with each installation — nobody reacts December 2 — exfiltration begins; tools updated Dec 2–3 More FireEye alerts fire as the exfiltration malware is upgraded December 12 — the Department of Justice notifies Target Target testified it was not aware of the breach until this contact December 15 / 19 — breach confirmed, most malware removed; public disclosure December 19: Target confirms about 40 million card accounts exposed January 10, 2014 — up to 70 million personal records also disclosed

What this shows: roughly a month inside the network before anyone outside Target noticed — and the two detection moments (around November 28 and November 30 – December 3) when Target's own software saw the attackers' tools and the alerts changed nothing.

At least two months before the breach — the key is copied. Attackers research Target's vendors using publicly available information (the report notes Target's supplier portal and facilities pages were findable by simple internet searches), then send Fazio Mechanical phishing emails carrying password-stealing malware. Fazio's free, consumer-grade anti-malware — with no real-time monitoring — does not stop it. The attackers now hold the contractor's credentials for Target's systems.

November 12 — the door opens. The attackers first gain access to Target's internal network. The states' later investigation put the same date on it precisely: on or about November 12, 2013, the attackers accessed Target's gateway server through credentials stolen from a third-party vendor, then exploited weaknesses to reach a customer service database and install malware. A stolen password with no second factor was enough.

November 15–28 — the registers are seeded, quietly. The attackers install their tailored BlackPOS malware on a small number of POS terminals first — a test. Around November 28, Target's Symantec antivirus software also detects malicious behaviour, implicating the same internal server FireEye would flag days later. Nothing stops the rollout.

November 30 – December 3 — the theft machinery goes in, alarms ringing. By November 30 the majority of Target's POS system is infected. The attackers also install a second kind of malware on a Target server, designed to move stolen data through the network and out through the firewall: three variants go in on November 30, updated twice more — just before midnight on December 2 and just after midnight on December 3. Per the report, FireEye triggered urgent alerts with each installation of this exfiltration malware (one alert carried the generic name "malware.binary"). Target's security team neither reacted to the alarms nor allowed FireEye to automatically delete the malware. The report fairly notes one possible reason: staff may have read the generic alert as a false positive on a system that alarmed often. That is an explanation of the failure, not an excuse for the design — an alert stream noisy enough to ignore is itself a finding.

December 2 onwards — the data walks out in daylight. From December 2, the malware sends stolen data by FTP (file transfer protocol, a standard way of moving files) through another compromised Target server to outside drop sites. Analysts at Dell SecureWorks, who examined a sample of the actual malware, found the exfiltration ran between 10:00 a.m. and 6:00 p.m. Central Standard Time — presumably, the report says, to hide inside Target's busier shopping hours. In total the attackers collected 11 GB of stolen information using a Russia-based server; other reporting described additional drop locations on compromised servers in Miami and Brazil. The card data itself covers cards used in U.S. Target stores between November 27 and December 18, 2013 (the window Target gave Congress; the states' investigation dates the malware's capture window November 27 to December 15).

How RAM scraping steals a card number CARD SWIPE customer pays at the register REGISTER MEMORY plaintext, briefly, before the processor BLACKPOS COPIES IT the malware reads the same memory OUT BY FTP via an internal staging server THE ALERT THAT CHANGED NOTHING FireEye flagged each installation of the exfiltration malware. Auto-delete was off, and no one followed up. The legitimate payment also flows from the same memory to the processor — that part worked normally, which is why nothing looked broken at the tills while the copying ran for over two weeks. Exfiltration ran 10:00 a.m.–6:00 p.m. Central — hidden inside the busiest shopping hours.

What this shows: the theft needed no broken encryption. Card data is briefly readable in the register's own memory; BlackPOS simply read the same memory, and a second piece of malware ferried the copies out — while the detection system watched and reported, unanswered.

December 12–19 — discovery comes from outside. Target testified that it was not aware of the breach until the Department of Justice contacted it on December 12. On December 15, per the Senate timeline, Target confirmed the breach and removed most of the malware. Independent reporter Brian Krebs broke the story on December 18, and on December 19 Target publicly confirmed that some 40 million credit and debit card accounts had been exposed. Stolen card data was already being sold on underground "card shops," complete — the report notes — with a seller nicknamed "Rescator" offering a money-back guarantee on quickly cancelled cards.

January 2014 — the second disclosure. On January 10, 2014, Target disclosed that non-financial personal information — names, addresses, phone numbers, and email addresses for up to 70 million customers — was also stolen. (The Senate report is candid that public reporting did not make clear exactly how the attackers reached this separate data set.) On January 12, Target's CEO confirmed publicly that malware on the POS terminals had enabled the card theft. Counting both sets together, the report's executive summary describes financial and personal information of as many as 110 million Target customers.

The damage

The people. About 40 million credit and debit card accounts were exposed — card numbers with the data from the magnetic stripe, including, reporting confirmed, encrypted PIN numbers for debit cards. Separately, personal information for up to 70 million customers was taken, and the states' 2017 settlement describes the combined picture as more than 41 million payment card accounts affected and contact information exposed for more than 60 million customers. Those records fed underground card shops within days, where fraudsters buy fresh card data at a premium precisely because banks have not yet had time to cancel the cards.

The consumer settlement. The lawsuits were consolidated in the U.S. District Court in Minnesota. The court preliminarily approved a consumer settlement on March 19, 2015 and granted final approval on November 17, 2015: Target paid $10 million into a settlement fund for class members, with people who could document their losses eligible for reimbursement of out-of-pocket loss and time (up to two hours at $10 per hour) up to a maximum of $10,000 per person; people without documentation were estimated at the time to receive about $40 each from what remained. Target also paid all notice and administration costs on top of the fund — the court's order records that notice reached more than 80 million people, with direct notice to 61 million — and paid $6.75 million in attorneys' fees separately. The settlement also required Target to improve its data security practices in significant ways.

The states' settlement. On May 23, 2017, 47 states and the District of Columbia settled their investigation for $18.5 million — at the time, the largest multistate data breach settlement ever achieved, led by the attorneys general of Connecticut and Illinois. That is $28.5 million in the two settlements alone, before Target's own investigation, remediation, and card-reissue costs, which these documents do not total — so this page does not invent a grand total.

The trust damage is the harder number. Target learned of one of the largest retail breaches in history not from its own monitoring — which had, in fact, detected the attackers' tools — but from the Department of Justice, a month after the intrusion began, in the middle of the Christmas shopping season. The Senate report's framing is the lasting one: this was not a breach that succeeded because no defence existed. The defences existed. The chain of small failures is the damage report.

How they recovered

Step one: confirm and clean, December 15. After the December 12 notification from the Department of Justice, Target confirmed the breach and, by December 15, had removed most of the malware from its network, per the Senate timeline. Public disclosure followed on December 19, with the fuller personal-data disclosure on January 10, 2014.

Step two: protect the customers whose data was out. Target offered identity-theft protection to consumers in the wake of the breach (noted in the federal court's settlement order), and — following a December 2013 recommendation from the New York Attorney General's office — agreed to provide free credit monitoring to potential victims. The 2017 multistate settlement also credits the previously negotiated credit monitoring for impacted consumers as part of the overall resolution.

Step three: settle, with security terms attached. Neither settlement was money-only. The 2015 consumer settlement required Target to improve its data security practices in significant ways, and the 2017 states' settlement wrote the control list into an enforceable agreement (next section). Recovery, in other words, ended with outside parties — a federal court and 47 state attorneys general — holding Target to the fixes.

The fix: technology and controls that came out of it

The 2017 multistate settlement is the cleanest statement of the fix, because it lists the controls Target agreed to maintain. Read it against the five weaknesses in "The problem" — each control answers one:

Segment the cardholder data. The settlement requires Target to segment its cardholder data environment from the rest of its computer network. This is the direct answer to the leaked compartments: a billing-system foothold should never have a network path to the registers. Segmentation means firewall rules and architecture that make the card zone reachable only by the systems that must reach it — and by nothing a heating contractor's login can touch.

Control access like it matters: rotation and a second factor. The agreement requires steps to control access to the network, including password rotation policies and two-factor authentication for certain accounts. Two-factor authentication for vendor and remote access is the direct answer to the stolen Fazio password: a copied password alone should open nothing.

A named, accountable security program. Target must develop, implement, and maintain a comprehensive information security program, employ an executive or officer responsible for executing the plan, and hire an independent, qualified third party to conduct a comprehensive security assessment. Governance is a control: someone senior owns the program, and someone outside the company checks it.

Encryption and maintained software. The settlement requires Target to maintain and support software on its network and to maintain appropriate encryption policies, particularly for cardholder and personal information data — shrinking the windows in which plaintext data like the registers' memory copy is the only copy an attacker needs.

And the lesson no settlement can write for you: answer the alarms. The Senate report's kill-chain analysis adds the operational half of the fix. FireEye detected the exfiltration malware as it was installed and again as it was updated, and decoded where the stolen data was being staged. Acting on that — or simply letting the tool delete what it found — could have stopped the theft at the delivery step, before a single card left. The durable controls are boring ones: alert triage with named owners and response times, automatic containment for high-confidence detections, and periodic proof — by test, not by assumption — that an urgent alert at 11 p.m. reaches a human who can act. Visa had also warned the industry about this exact malware class twice in 2013; threat warnings only work if somebody's job is to read them and change a setting.

What this means for you

Enterprise network engineers

This is your case study, because none of it required a hyperscaler's budget to prevent. Your vendor list is your attack surface: every contractor, MSP, and facilities supplier with remote access is a door, and the door is exactly as strong as their security — Fazio's free consumer anti-malware was, in effect, part of Target's perimeter. Inventory vendor access, require two-factor authentication for all of it without a "low-level contractor" exception, and give each vendor the smallest possible reach. Then segment: draw your cardholder, customer-data, and crown-jewel zones, and prove with firewall rules — not diagrams — that a billing portal cannot route to them. Packet Path's free tools below practise both halves of that check. Finally, audit your defaults: the reported BMC default-account theory is unproven for Target, but default and unused accounts are free for you to eliminate today.

Hyperscaler engineers

Your version of Fazio is the third-party SaaS integration, the support vendor with a break-glass account, the CI service with a token that never rotates. The controls scale but the failure shape is identical: a credential minted for a narrow job, usable from anywhere, with no second factor and no anomaly check on where it is used from. The Senate report makes a point worth stealing for your own reviews: watch where a credential is used, not just whether it is valid — a vendor login appearing in an unrelated part of the network is a detection, if anyone is looking. And take the FireEye lesson at platform scale: detection coverage is not a metric. Time-to-human-action on a high-confidence alert is. If your auto-remediation is switched off because the tool is noisy, the finding is the noise, not the switch.

HFT / low-latency engineers

Your registers are your order gateways and market-data handlers, and your plaintext window is the unencrypted feed or session inside your own perimeter. The transferable points are segmentation and blast radius: the systems that can place orders should be reachable from the fewest possible networks, vendor and support access into a trading environment should be jump-hosted, recorded, and two-factored, and a compromise of a back-office system should have no route — none — to a matching-engine segment. Target also shows why "we are compliant" is a weak sentence in a post-mortem: it was PCI DSS certified in September 2013 and breached in November. Certification is a snapshot; the kill chain runs every day.

Try it yourself

Two free Packet Path tools practise the two checks that would have broken this chain. The ACL Analyzer reviews firewall and router access lists — the rules that are segmentation in practice — and flags shadowed, redundant, and over-broad rules, so you can see whether your "isolated" zone is actually isolated. The Network Config Verifier compares device configurations side by side and flags mismatches and risky leftovers — the habit that catches default accounts, stale vendor access, and drift between devices that are supposed to match. Run both against your own edge: the Target breach is what "we never checked" looks like at national scale.

Sources

Primary source — U.S. Senate Committee on Commerce, Science, and Transportation, Majority Staff Report for Chairman Rockefeller, A "Kill Chain" Analysis of the 2013 Target Data Breach (March 26, 2014). The incident narrative, the Fazio Mechanical entry, the BlackPOS/RAM-scraping mechanics, the November 15–30 installation windows, the FireEye and Symantec detections, the December 2 FTP exfiltration and 11 GB figure, the December 12 DOJ notification, and the kill-chain analysis are taken from this report — which states that it is itself based on media reports and expert analyses published since December 19, 2013, and flags the internal-movement step and the BMC default-account route as not fully established. https://www.commerce.senate.gov/index.cfm?a=files.serve&File_id=24d3c229-4f2f-405d-b8db-a3a67f183883

Court record — U.S. District Court, District of Minnesota, In re: Target Corporation Customer Data Security Breach Litigation, MDL No. 14-2522 (PAM/JJK), Document 645, Memorandum and Order granting final approval of the consumer class settlement (filed November 17, 2015): the $10 million settlement fund, documented-loss payments up to $10,000, the estimated ~$40 undocumented payment, notice reaching more than 80 million people with direct notice to 61 million, and the $6.75 million attorneys' fee award paid separately. https://www.govinfo.gov/content/pkg/USCOURTS-mnd-0_14-md-02522/pdf/USCOURTS-mnd-0_14-md-02522-6.pdf

Settlement — Office of the Attorney General of New York, "A.G. Schneiderman Announces $18.5 Million Multi-State Settlement With Target Corporation Over 2013 Data Breach" (May 23, 2017), announcing the settlement of 47 states and the District of Columbia (the largest multistate data breach settlement to that date): the states' findings (gateway server accessed on or about November 12, 2013 through third-party vendor credentials; more than 41 million payment card accounts and contact information for more than 60 million customers) and the required controls (comprehensive information security program, a responsible executive or officer, an independent third-party security assessment, encryption policies, segmentation of cardholder data, password rotation, and two-factor authentication for certain accounts). https://ag.ny.gov/press-release/2017/ag-schneiderman-announces-185-million-multi-state-settlement-target-corporation

No other sources are used for facts on this page. Where the Senate report and the states' investigation give slightly different windows or totals, both are stated and attributed in the text.

Key takeaways

Keep reading