Encryption: who bears the cost when access is hijacked?

The economic cost of a backdoor · Part 5 of 7

An intruder can turn a network’s own capabilities against it. From US telecom breaches and a Greek precedent to French insurance claims and Chinese security rules, this investigation follows the damage and the question of who pays.

On 25 July 2025, Orange detected an intrusion into one of its information systems. The telecoms operator isolated services that might have been affected. Three days later, it said this protective action had itself disrupted some services and management platforms, mainly in France. At that stage, investigators had found no evidence suggesting that data had been exfiltrated. The statement put no figure on the damage. ^V12

This sequence offers a useful starting point for the economics of an intrusion. A company may have to suspend operations to keep an attack from spreading. It then commits resources to making its systems safe to use again, while customers try to keep their own work moving. Sometimes, the cost begins with the decision to close a connection.

The Orange statement documents that containment mechanism. It establishes neither a backdoor nor a connection to Salt Typhoon. For the encryption debate, the next question is more specific: when a system includes a particular power to read data or exercise control, how far can the misuse of that power extend the consequences of an intrusion?

US investigations, technical research and French and Chinese publications help explain how incidents unfold and how their costs are shared. They also leave a substantial gap. Within the evidence assembled for this investigation, the total economic cost attributable to the misuse of an exceptional-access capability remains unknown.

The capabilities an intruder reaches

A backdoor can mean a mechanism introduced without the owner’s knowledge to bypass security. In the public debate over encryption, the same word is also used to criticise access sought by public authorities. To follow the evidence, we need to distinguish exceptional access, designed to make certain data readable under specified conditions, from the administrative functions that keep a network running. They serve different purposes. Misuse of either can give an intruder greater powers, with the reach determined by the particular system. ^V20

In April 2025, the US Federal Bureau of Investigation (FBI) described the results of a campaign publicly known as Salt Typhoon, which it attributed to actors affiliated with China: stolen call logs, access to private communications involving a limited number of identified victims, and copies of selected information subject to US court-ordered law-enforcement requests. That last category draws attention to information associated with lawful interception. It identifies what was obtained. The statement leaves the initial entry point unresolved. ^V01

Cisco Talos supplied another part of the picture. In its February 2025 analysis, the vendor reported evidence suggesting that a vulnerability had probably been exploited in one case. In the other cases it investigated, initial access to Cisco devices relied on legitimate credentials. How those credentials were first acquired remained unknown. ^V03

The evidence has continued to develop. Talos’s report published on 29 October 2026, covering its third-quarter incident-response engagements, describes prolonged telecoms intrusions: valid accounts, a vulnerable management interface, then legitimate network functions used to maintain access or collect data. One campaign deployed an implant known to be used by Salt Typhoon. That technical connection is insufficient to attribute every engagement during the quarter to the same group. ^V04

The sequence matters. A traffic-copying function may become useful after an attacker enters a network. A database containing information related to court orders may be reached without having provided the way in. Collapsing those stages obscures the central economic question: what additional capability did the attacker obtain, and what consequences did it make possible?

A joint advisory from several US agencies and international partners was published on 27 August 2025 and revised on 3 September as version 1.1. It describes movement between devices and networks through trusted connections. Its scope only partially overlaps with the groups named by vendors, including Salt Typhoon. It brings together technical findings; the geopolitical attribution belongs to the issuing authorities. ^V02

China disputes that attribution. In a response published on 1 September 2025, its foreign ministry also repeated an account involving an American “backdoor”, explicitly attributing that explanation to US media. The statement records a diplomatic position. It supplies no independent technical verification of the route taken by the intruders. ^V08

Method and scope

Qualitative diagram. Greece, 2004–2005: the investigation published in 2007 describes unauthorised code using an interception function present in the switches; the initial entry remains unresolved in that reconstruction. Dotted lines mark this limit. Telecom publications from February 2025, September 2025 and October 2026 report several intrusions involving valid access, vulnerabilities and trust relationships. They document subsequent use of network functions to collect traffic. The diagram combines neither cases nor attributions and establishes no common cryptographic entry route. Historical Greek telephony and modern end-to-end encrypted messaging have different architectures.

An authorised function becomes a covert power

An earlier case gives us a more direct view of an interception capability being misused. In Greece in 2004–2005, unauthorised code installed on switches in Vodafone’s network used a wiretapping function already present in their software. A 2007 investigation by Vassilis Prevelakis and Diomidis Spinellis identified an unusual arrangement: the operator had the technical interception capability but had not purchased its usual management interface. The added code activated the capability to copy communications. The initial point of entry remained unresolved in their reconstruction. ^V17

This was a historical telephone-network architecture, different from a modern messaging service protected by end-to-end encryption. Its relevance lies in the distinction between getting into a system and the power awaiting an attacker there. A capability created for authorised use can expand what a successful intrusion allows.

The problem disclosed at Juniper in December 2015 took a different form. CERT/CC’s notice distinguishes two independent ScreenOS vulnerabilities: one allowed unauthorised administrative access; the other made it possible to decrypt observed VPN traffic. A VPN creates an encrypted tunnel between points on a network. The vulnerabilities therefore affected different security functions. ^V05

Researchers examining the second mechanism in a paper presented at ACM CCS in 2016 explained the role of Dual EC, a generator of numbers used in cryptographic operations. Using a real device and their own parameters, they demonstrated that a VPN session could be decrypted passively. Their experiment established a technical capability. It provided no count of actual victims or losses and did not identify whoever introduced the unauthorised code. ^V06

These cases show why “backdoors” cannot be treated as a single object. A misused interception function, covert administrator access and a weakness that enables decryption have different requirements and reach. Any assessment of their cost has to begin by identifying the mechanism involved.

The reach of a key

Imagine two organisations holding the same files. In the first, one privileged capability opens several sets of data. In the second, those sets rely on genuinely separate keys and controls. If an attacker compromises just one of these capabilities, the possible consequences differ, provided the separation also holds at the level of administration and recovery.

The US National Institute of Standards and Technology (NIST) expresses this principle in its key-management recommendations: design a system so that compromising one key exposes as little information as possible. Preparations should also cover revocation, replacement and recovery. This offers a way to limit exposure. It supplies no universal percentage reduction in risk. ^V19

The idea of a “key to everything” becomes misleading when it takes the place of examining an architecture. Exceptional access might be restricted to one device, one account or a particular period. Conversely, several supposedly separate keys may all depend on the same administrator account. In the illustration below, independence is a demanding assumption: the compromised access must actually stop at the boundary shown.

Method and scope

Conceptual comparison informed by NIST SP 800-57 Part 1, revision 5, May 2020, section 5.5.2. Letters A to D identify illustrative datasets, without real quantities. Rose marks a control assumed to be compromised and the potentially accessible perimeter in this model. Independence requires genuinely separate accounts, permissions, administration paths and recovery mechanisms. Gray controls are not an observed guarantee. The drawing measures neither compromise probability nor expected annual loss. Each architecture’s safeguards could also change entry probability, which requires separate assessment.

For an insurer, this reasoning changes the scale of the problem. When several clients depend on the same provider or technology, a common event can affect multiple policyholders. EIOPA, the European insurance supervisor, discusses this accumulation risk in its December 2025 Financial Stability Report. It also observes that the still-limited size of cyber insurance reduced its direct impact on insurers’ stability at that time. ^V14

Both dimensions matter. Shared control can extend the consequences of a successful compromise, while the protections around it can change the likelihood of success. Comparing two architectures would require assessing both. Mapping the reach of an access capability tells us about potential exposure; calculating an expected annual loss takes further evidence.

The French experience calls for another distinction. In its review of 2025, published in March 2026, ANSSI describes activity associated with Salt Typhoon and a compromised device used as a relay. The agency also says the widespread deployment of such equipment makes it difficult to distinguish strategic targeting of telecoms from broader opportunism. Infrastructure used along the way and the target ultimately sought play different roles in an investigation. ^V11

Restoring systems after an intrusion

After detection, those responsible have to decide what to isolate, which access to suspend and which backups they can trust. NIST distinguishes containment, which aims to halt the spread of an incident, from eradication of the means of persistence and recovery. Bringing services back into production includes checking backups and restored systems. Some sophisticated intrusions whose full extent remains unclear may require hardware replacement. ^V18

This is where security uses tangible resources: staff time, replacement capacity, testing and user support. If a service remains unavailable during the work, the loss also depends on how much activity can be deferred, recovered or continued through other means. An hour of interrupted operations does not automatically become an hour of revenue lost for good.

France’s cybersecurity agency, ANSSI, recommends preparing remediation before a crisis: the work needed to restore a trustworthy system. Its January 2026 guide covers backups, communications that can be used during an incident and the activation of insurance services under the relevant contractual terms. Preparation is an upfront commitment of resources so that the organisation can act when some of its usual tools become suspect. ^V15

A different limit emerges once readable information has been copied outside the system. Replacing a key can secure future exchanges. It cannot erase a copy of plaintext already held by an intruder. Service may be restored while the confidentiality of that information remains lost. This follows from how copying works, regardless of the amount spent rebuilding the system.

Method and scope

An original explanatory model, without estimated duration or cost. The sequence simplifies incident-response functions; stages can be revisited. Orange’s statement of 28 July 2025 describes isolation-related disruption after detection on 25 July. At that stage, it reports no evidence indicating exfiltration. The rose branch is a general scenario involving a readable copy already held outside, separate from that incident. It attributes neither a backdoor nor a Salt Typhoon link to Orange. Revoking or replacing access protects according to the architecture; a plaintext copy already obtained has a separate fate. Guidance sources: NIST, April 2025, and ANSSI, January 2026.

A lost trade secret might, for example, change the outcome of a future negotiation. That is a possible route through which an intrusion could affect economic outcomes; it is not an observed loss in the preceding cases. Assigning it a value would require knowing what was taken, how it was used and what would have happened without the intrusion. Repair costs may be documented while that comparison remains out of reach.

China’s reporting requirements highlight the time needed to establish the facts. Rules in force since 1 November 2025 allow an initial report with incomplete information, followed by updates as the investigation develops. A summary report must be submitted within thirty days after incident handling has ended. The deadline is measured from completion of the response, rather than detection. ^V09

A separate regulation, in force since August 2025 for designated critical information infrastructure within its scope, requires staff and budgets for commercial cryptography, along with a security assessment at least annually. It identifies tasks required by regulation. Actual expenditure and the effectiveness achieved need to be established separately. ^V10

The scope of insurance data

An economic investigation benefits from observed amounts. In France, the French corporate risk-management association AMRAE’s LUCY 2026 study brings together 2025 data from twelve brokers and SMABTP. It records 1,251 claims and a reported indemnification amount of €83,203,654 within that coverage. The definition refers to amounts reported during the year; the publication provides no breakdown that would allow this total to be treated as cash already paid. ^V13

Companies with annual revenue above €1 billion account for 63 claims and €44,583,393: 5.0% of the count and 53.6% of the amount. These shares are calculated from the same table. They show that amounts are heavily concentrated in this category of policyholders. ^V13

This concentration helps explain why counting incidents leaves much of the financial risk out of view. Several small claims and one large event can produce very different financial outcomes. Establishing whether companies were affected together by a common failure would, however, require the dates, causes and technical dependencies behind each claim. LUCY’s breakdown by company size does not establish that correlation.

The scope imposes other limits. Coverage of small and microbusinesses is indicative, and local authorities are excluded. The surveyed portfolio changes, as do some deductibles. This investigation therefore uses the 2025 table without treating the comparison with 2024 as a change measured on a constant sample. LUCY also does not identify claims attributable to a backdoor. ^V13

There is a useful methodological lesson here: a meaningful figure must retain its denominator. The claim count tells us how many claims appear in the observed portfolio. The reported amount describes another dimension. Neither figure alone measures the total damage to the French economy.

Following the loss to whoever bears it

Imagine a company funding the rebuilding of a service after an intrusion. Its specialists and contractors devote work to the recovery; some equipment is replaced; some business may be lost. Those resources and consequences form part of the cost. An insurer may subsequently indemnify covered expenses or losses under the policy.

Payment then changes who finances the burden. Adding the same repair expense and the indemnity that reimburses it would count that cost twice when estimating total damage. Assessing the claim, administering it and dealing with additional consequences may use further, distinct resources. Those need to be identified separately. This basic accounting distinction helps trace the flows without assuming the terms of any particular policy.

The contract also determines what the company retains. A deductible leaves it bearing part of the risk; a limit caps the possible payment; the scope of cover determines which events and losses qualify. EIOPA also discusses implicit cyber exposure in policies whose treatment of that risk is insufficiently clear. Analysis therefore has to examine the provisions that actually apply. ^V14

Customers may face downtime, workarounds or the exposure of their data. Depending on their contracts and available remedies, they may retain some of that burden. This can create a gap between the loss visible in the supplier’s accounts and the consequences borne by others around it. Those dependencies need examining before a total can be estimated.

The risk of hijacked access is therefore also a question of how losses are distributed. The party that designs or requires the capability, the operator that runs it and the people whose data become accessible may all be different. An implementation budget answers a question about initial funding. A fuller economic assessment also needs to establish who bears recovery costs, interruptions and the consequences of disclosure.

The trade-off needs an architecture and evidence

The strongest argument for exceptional access concerns investigations that run into unreadable data. The National Academies recognise this difficulty in their 2018 analytical framework. They also identify the additional complexity and risk introduced by such access, tying their scale to the way a system is designed and operated. The available evidence limits the quantification of both benefits and harms. ^V20

That calls for examining a specific design. What data would it make readable, following whose authorisation, with which controls? What could an actor obtain by subverting those controls? A capability confined to one device and one shared across an entire fleet require different assessments. Proposed safeguards belong in the comparison, alongside their cost and the investigative opportunities they create.

End-to-end encryption serves a specific purpose within this picture: limiting who can read messages as they travel between participants. Version 2.0 of the US Cybersecurity and Infrastructure Security Agency’s (CISA) mobile guidance, dated 24 November 2025, added advice concerning these encrypted messaging services. Device and account security remain separate questions: a message displayed on a compromised device may be accessible to the intruder. ^V07

Evidence that challenges a narrative of automatic escalation also deserves attention. In its June 2026 report, the Banque de France says the French authorities had identified no increase in cyberattacks on the French financial sector resulting from the geopolitical tensions examined during the first half of the year. This finding is confined to that relationship, period and sector. It illustrates the value of distinguishing an elevated threat from its measured occurrence. ^V16

The public documents assembled here establish several points: powerful functions can be misused; the reach of access shapes the possible consequences; and responding to an intrusion requires resources and can disrupt activity. The economic total attributable to a particular exceptional-access design remains open.

A decisive next step would be an assessment connecting the expected investigative benefit, the reach of the access created and the additional losses that access could make possible. That would let readers examine the assumptions, compare safeguards and follow the burden to those who bear it. Such evidence would turn the price of a backdoor into an economic question that can be tested.

---

Method. Documentary investigation closed on 7 November 2026, using twenty primary references: official texts, original technical publications, company documents and an industry study. Attack attributions remain those of the cited authors. LUCY calculations are ratios derived from the 2025 table; the other illustrations explain conditional mechanisms. Chinese regulations are presented from official Chinese-language publications; translations are editorial. No interviews, infrastructure audits or original estimates of damage were conducted. The references give dates, passages used and the scope of the documents.

Further reading

The sixth part on alternative routes to evidence examines available sources, investigative resources and the work of establishing evidence.

The first part on Parliament’s proposed encryption safeguard examines the French proposal; the second on devices and backups follows copies of a business secret. The third on state access examines oversight, while the fourth on iCloud and borders explores safeguards and switching costs.

The analysis of cyberinsurance and a customer’s outage examines the scope of contracts following an interruption.

The sharing image depicts a repaired network and document copies already outside its perimeter. The conceptual composition illustrates a conditional scenario; it depicts no actual incident or observed decryption.

Sources

1. V01 · Federal Bureau of Investigation · 2025-04-24 · Accessed 7 November 2026FBI Seeking Tips about PRC Targeting of U.S. Telecommunications
   Passages and scope

   FBI/IC3 alert of 24 April 2025, opening paragraphs: the agency attributes the campaign to PRC-affiliated actors and reports stolen call logs, a limited number of private communications involving identified victims, and copies of selected information subject to court-ordered requests. These are reported outcomes and an official attribution. The alert establishes neither an entry through lawful-interception systems nor a breach of end-to-end encryption, and gives no economic loss estimate.

   Additional primary document

2. V02 · NSA, CISA, FBI, DC3 et partenaires internationaux · 2025-08-27 · Accessed 7 November 2026Countering Chinese State-Sponsored Actors Compromise of Networks Worldwide to Feed Global Espionage System
   Passages and scope

   Joint advisory first published on 27 August 2025 and revised on 3 September as version 1.1, according to its history on printed page 30. The summary and technical sections describe compromised devices, trusted connections, persistence and collection. The activity only partially overlaps commercial group names including Salt Typhoon. Attribution belongs to the signatory authorities; the observed paths and remediation advice supply no measured incident cost.

   Additional primary document

3. V03 · Cisco Talos · 2025-02-20 · Accessed 7 November 2026Weathering the storm: In the midst of a Typhoon
   Passages and scope

   Original Cisco Talos analysis of 20 February 2025, summary and initial-access sections: one investigated case suggests likely exploitation of CVE-2018-0171; the other initial accesses to Cisco devices examined by that date used legitimate credentials. How the initial credentials were obtained remains unknown. Capture and management functions appear in subsequent operations. These findings concern the vendor’s investigations rather than every intrusion worldwide.

4. V04 · Cisco Talos : Aliza Johnson · 2026-10-29 · Accessed 7 November 2026IR trends Q3 2026: Actors impersonate IT help desks in vishing attacks as ransomware activity increases
   Passages and scope

   Talos incident-response findings for the third quarter of 2026, published on 29 October. The cyberespionage section describes valid accounts, a management-interface vulnerability and the misuse of legitimate network features. One campaign used BadCandy, an implant known to be used by Salt Typhoon. That technical link does not attribute all engagements in the quarter to the same group. The report publishes no loss total for these cases.

5. V05 · CERT Coordination Center · 2015-12-21 · Accessed 7 November 2026Juniper ScreenOS contains multiple vulnerabilities : VU#640184
   Passages and scope

   CERT/CC vulnerability note VU#640184, published on 21 December 2015 and revised the following day. CVE-2015-7755 concerns unauthorized remote administration; the independent CVE-2015-7756 can enable decryption of observed VPN traffic under its technical conditions. The note documents affected versions and fixes. It identifies neither the author of the unauthorized code nor a victim count or measured economic loss.

6. V06 · Stephen Checkoway et neuf coauteurs : ACM CCS 2016 ; copie des auteurs · 2016-10 · Accessed 7 November 2026A Systematic Analysis of the Juniper Dual EC Incident
   Passages and scope

   Original analysis by Checkoway and co-authors, ACM CCS, October 2016. The abstract, introduction and experiment explain how a Dual EC parameter can enable passive decryption, demonstrated on a NetScreen device using the researchers’ own parameters. This establishes a technical capability in a particular implementation. It is separate from observations of real-world stolen data, victim numbers, losses or the identity of whoever introduced the unauthorized code.

   Additional primary document

   Additional primary document

7. V07 · Cybersecurity and Infrastructure Security Agency ; copie préservée par U.S. Government Publishing Office · 2025-11-24 · Accessed 7 November 2026Mobile Communications Best Practice Guidance : Version 2.0
   Passages and scope

   CISA guidance, version 2.0 dated 24 November 2025: version history and introduction on the first page, general recommendations on the second. The update expands the especially targeted audiences and adds protections for end-to-end encrypted messaging, including account and linked-device precautions. This dated guide provides recommendations. Protection of communications in transit is separate from the security of a compromised device or account.

   Additional primary document

   Additional primary document

8. V08 · Ministère des Affaires étrangères de la République populaire de Chine · 2025-09-01 · Accessed 7 November 2026Foreign Ministry Spokesperson’s Remarks on the US Ganging up with Allies to Hype Up ‘Salt Typhoon’ Hacking Operations
   Passages and scope

   Chinese Foreign Ministry response published on 1 September 2025: rejection of Western allegations and an assertion about an American backdoor being exploited, explicitly attributed to American media reporting. This is a primary source for China’s diplomatic position. It supplies no independent technical investigation establishing the initial intrusion route, decryption capability or economic losses.

   Additional primary document

9. V09 · Cyberspace Administration of China (CAC) · 2025-09-15 · Accessed 7 November 2026Measures for the Reporting of Cybersecurity Incidents
   Passages and scope

   Articles 2, 4 to 8, 11 to 14 and the annex. Issued on 11 September 2025, published on 15 September and effective from 1 November. Initial reporting deadlines depend on incident severity and operator type. Article 7 allows limited initial information followed by updates; Article 8 requires a summary within thirty days after handling ends. Direct-loss thresholds classify incidents rather than measure observed damage.

10. V10 · Administration nationale de la cryptographie, CAC et ministère chinois de la Sécurité publique · 2025-06-27 · Accessed 7 November 2026Regulations on the Use of Commercial Cryptography in Critical Information Infrastructure
    Passages and scope

    Joint Order No. 5, dated 11 June 2025, published on 27 June and effective from 1 August. Articles 2, 7 to 9 and 12 to 18 cover designated critical infrastructure, skills, budgets, pre-operation checks and at least annual assessments. Written requirements establish neither an operator audit nor actual technical independence between roles. Required budgets are separate from expenditure incurred.

    Additional primary document

11. V11 · ANSSI / CERT-FR · 2026-03-11 · Accessed 7 November 2026Cyber Threat Overview 2025 : CERTFR-2026-CTI-002
    Passages and scope

    ANSSI’s 2025 threat panorama, published on 11 March 2026. Printed page 18 describes broader victim targeting associated with Salt Typhoon in France and a compromised device used as a pivot. ANSSI says it cannot determine whether strategic telecom targeting or wider opportunism involving common equipment predominates. Pages 48 to 50 discuss suppliers and their customers. These observations establish no connection between the Orange incident and Salt Typhoon.

    Additional primary document

    Additional primary document

12. V12 · Orange Newsroom · 2025-07-28 · Accessed 7 November 2026Orange announces a complaint filed on 28 July following an intrusion into one of its information systems
    Passages and scope

    Orange statement of 28 July 2025: detection on 25 July, isolation of potentially affected services, and disruption to some services and management platforms, mainly in France. At the statement date, the company reported no evidence suggesting exfiltration. Reopening by 30 July was a forecast. This provisional company account establishes neither a backdoor, an attacker’s identity nor a quantified loss.

13. V13 · AMRAE · 2026-05-29 · Accessed 7 November 2026LUCY : A Spotlight on Cyber Insurance, 2026 edition
    Passages and scope

    LUCY 2026, reporting year 2025, methodology and claims table: twelve brokers and SMABTP, 20,996 policies, 1,251 claims and EUR 83,203,654 in reported indemnification. Companies with revenue above EUR 1 billion account for 63 claims and EUR 44,583,393. Calculated shares are 5.0% and 53.6%. The insured sample, limited small-business coverage and excluded public bodies cannot establish total French economic losses or the cost of backdoors.

14. V14 · EIOPA · 2025-12-15 · Accessed 7 November 2026Financial Stability Report : December 2025, EIOPA-BoS-25-579
    Passages and scope

    EIOPA report of 15 December 2025, Chapter 4, pages 54 to 60. Dependence on a common service or technology can produce loss accumulation. The report also discusses silent cyber cover and notes that the line’s small relative size then limits its direct effect on insurers’ financial stability. Supervisory analysis and third-party sources have distinct roles; the report estimates no loss caused by exceptional access.

    Additional primary document

15. V15 · ANSSI · 2026-01-16 · Accessed 7 November 2026Cyberattacks and Remediation : Preparing for Remediation, Version 1.0
    Passages and scope

    ANSSI guide, version 1.0 of January 2026, publication notice dated 16 January. Pages 7 to 8, 17 to 18, 29 to 30 and 33 cover remediation planning, restoration, crisis communications and insurance services subject to contractual procedures and deadlines. It describes resources and preparations, without a mean cost, guaranteed recovery time or ability to retrieve copies already taken by an intruder.

    Additional primary document

    Additional primary document

16. V16 · Banque de France · 2026-06-24 · Accessed 7 November 2026Financial Stability Report : June 2026
    Passages and scope

    June 2026 report, published on 24 June, overview of geopolitical risk. The Banque de France and ACPR identify no increase in cyberattacks on France’s financial sector attributable to the tensions examined in the first half. The page’s 24 September update does not extend that observation period. Threat level, observed incidents and causal attribution remain separate assessments.

17. V17 · Vassilis Prevelakis et Diomidis Spinellis : IEEE Spectrum, 44(7), 26–33 · 2007-07 · Accessed 7 November 2026The Athens Affair
    Passages and scope

    Original investigation by Prevelakis and Spinellis, published in IEEE Spectrum in July 2007, concerning Greek wiretapping in 2004 and 2005. It describes unauthorized code, interception functionality already present in the switches, and management software the operator had not purchased. The code subverted an existing capability; the initial entry and perpetrators remained unresolved in this reconstruction. This historical telephone architecture differs from modern end-to-end encrypted messaging.

    Additional primary document

18. V18 · NIST : Alexander Nelson, Sanjay Rekhi, Murugiah Souppaya et Karen Scarfone · 2025-04-03 · Accessed 7 November 2026Incident Response Recommendations and Considerations for Cybersecurity Risk Management: A CSF 2.0 Community Profile : NIST SP 800-61 Rev.3
    Passages and scope

    Final NIST SP 800-61 revision 3, issued on 3 April 2025, printed pages 32 to 34: containment, eradication and recovery; RC.RP-03 and RC.RP-05 address backup and restored-asset integrity. Replacing hardware is contemplated for some sophisticated intrusions whose full scope remains unknown. These recommendations supply neither incident frequency nor a recovery duration or cost.

    Additional primary document

19. V19 · NIST : Elaine Barker · 2020-05 · Accessed 7 November 2026Recommendation for Key Management: Part 1 : General, NIST SP 800-57 Part 1 Rev.5
    Passages and scope

    Final NIST SP 800-57 Part 1 revision 5, May 2020, section 5.5.2, printed pages 50 to 52. System design should limit the data exposed by one compromised key and prepare revocation, replacement and recovery. The initial revision 6 draft published on 5 December 2025 retains draft status. Key separation, administration and recovery paths require joint assessment; no numerical risk reduction is supplied.

    Additional primary document

20. V20 · National Academies of Sciences, Engineering, and Medicine : consensus study report · 2018 · Accessed 7 November 2026Decrypting the Encryption Debate: A Framework for Decision Makers
    Passages and scope

    National Academies consensus report of 2018, Chapter 7, Boxes 7.1 and 7.2, pages 84 to 89. Its framework examines investigative difficulties, additional risks from exceptional access and limits to quantifying them. Effects depend on the mechanism’s technical and operational details. It supports an evaluation method rather than a current monetary total or a universal policy prescription.

    Additional primary document