Multipoint Caller Authentication - An AI based voice biometrics solution for real-time fraud detection and prevention in the call centre

英国的呼叫中心行业雇佣了大约81.2万人\[来源：仙人掌搜索\]每年处理超过10亿个电话，其中68%使用语音通道\[来源：呼叫中心助手2021\]。这个专注于支持消费者的行业，共同拥有英国大多数(如果不是所有)成年人的个人记录，使呼叫中心成为试图接管账户和其他欺诈性交易的网络犯罪分子的首要目标。**网络犯罪-社会工程**由于IT行业投资于安全强化，我们现在可以看到，攻击者已经重新专注于成为使用心理技巧的专家，例如使用情绪触发词，以影响人类的行为和漏洞，以创造信任感。鼓励受害者披露敏感信息的紧迫性和权威性。人为行为对企业和个人来说都是一个严重的脆弱性，82%的企业违规行为涉及人为因素和社会工程。\[来源：AAG 2022\]**呼叫中心的业务挑战**\*竞争格局--**需要在竞争激烈的呼叫中心业务中生存\*客户期望--**客户对无摩擦服务的期望\*员工流失--**需要持续意识培训的员工高流动率\*全球交付中心--**服务从多个中心(英国国内外)提供服务\*员工激励--**低薪酬和绩效压力导致士气低落**当前的方式努力防止欺诈**\*限制信息访问--**仅显示工程师需要查看的内容，这通常会导致与呼叫者共享看似无害的信息时的自满情绪\*意识培训--**处于性能压力下的工程师被要求识别可能的威胁并发出警报。\*监督--**对策略遵从性的监督非常少，以查看潜在的安全问题是否得到有效管理，因为只有不到2%的呼叫对话被检查。**CXReview建议的解决方案：多点呼叫者身份验证(MCA)**完全无摩擦，独立和被动的解决方案，可以在半天内部署在任何呼叫中心，开始监听网络犯罪分子。人工智能和ML驱动的呼叫监听器将在交互式语音响应(IVR)被触发后立即启动，以确定呼叫者是否需要通过额外的安全检查，使用音频信号分析来识别呼叫者(使用语音生物识别)。继续使用已知特征的人工智能模型和情绪特征标记来构建呼叫者行为档案，以构建罪犯的最完整图景。

The call centre industry in the UK employs about 812k people \[source: Cactus Search\] handling over a billion calls per annum, 68% of which are using the voice channel \[Source: Call Centre Helper 2021\]. This industry, which is focused on supporting consumers, collectively hold personal records of most, if not, all of the adult population in the UK, making call centres a prime target for Cyber criminals attempting to takeover accounts and other fraudulent transactions.**Cybercrime - Social Engineering**Since IT industry has invested in security hardening, we can now see that attackers have refocused to become experts using psychological techniques, such as using emotional trigger words, to influence human behaviour and vulnerabilities to create a sense of trust, urgency and authority to encourage the victims to disclose sensitive information.The human actions represent a critical vulnerability for both businesses and individuals with 82% of breaches against businesses involving a human element and social engineering. \[Source: AAG 2022\]**Call centres' business challenge**\***Competitive landscape --** need to survive in highly competitive call centre business\***Customer expectations --** customer expectations of frictionless service\***Staff attrition --** high turnover of staff that need continuous awareness training\***Global delivery centres --** services are delivered from centres from multiple centres (inside and outside UK)\***Staff motivation --** Low pay and pressures on performance leads to low morale**Current way of working to prevent fraud**\***Restrict information access --** Show only what agent needs to see, which often leads to complacency in sharing seemingly innocuous information with callers\***Awareness Training --** Agents under performance pressures are asked to identify possible threats and raise alerts.\***Oversight --** very little oversight on policy adherence to see whether potential security issues are managed effectively, since less than 2% of the call conversations are ever checked.**CXReview Proposed Solution: Multipoint Caller Authentication (MCA)**A totally frictionless, standalone and passive solution that can be deployed in any call centre within half a day to start listening for Cyber criminals.An AI and ML driven call listener will start as soon as the Interactive Voice Response (IVR) is triggered to ascertain whether the caller needs to go through additional security checks, using audio signal analysis to identify the caller (using Voice Biometric).Continually build the caller behavioural profile using AI models of known characteristic augmented with emotional characteristics markers to build the most complete picture of the criminal.