350-701
Free 350-701 practice questions
10 questions from our 350-701 practice tests, each with the correct answer and a written explanation. Read them with the answers shown, or switch to quiz mode and answer first.
Question 1 of 10, Security Concepts
An internal assistant built on a large language model summarizes web pages that employees submit. A tester shows that hidden text on one page makes the assistant ignore its instructions and reveal internal ticket data. Which AI/LLM vulnerability does this demonstrate?
- AIndirect prompt injection through untrusted content that the model processes Correct answer
- BVector and embedding weakness caused by poisoned entries in the retrieval database
- CSystem prompt leakage caused by storing secrets inside the model's system prompt
- DSupply chain compromise caused by a tampered pre-trained model from a public hub
Answer and explanation: The malicious instructions arrive inside external content the model ingests, which is indirect prompt injection; defenses include treating retrieved content as untrusted, constraining tool access, and output filtering. System prompt leakage is about the model disclosing its own system prompt. Embedding weaknesses involve the vector store itself, which is not used here. A supply chain issue would involve a compromised model or dependency, not a submitted web page.
Question 2 of 10, Security Concepts
Authentication logs show thousands of login attempts against many different accounts from hundreds of IP addresses, each account tried only once or twice with a username and password pair that looks real. Which attack is this, and which control MOST directly reduces its success?
- ABrute force; lock each account after three failed login attempts within one hour
- BCredential stuffing; enforce MFA and screen passwords against known breach lists Correct answer
- CPass-the-hash; disable NTLM authentication on all internal file servers
- DPassword spraying; require password changes for all users every thirty days
Answer and explanation: Reused credential pairs from earlier breaches tried across many accounts and sources is credential stuffing, and MFA plus breached-password screening makes stolen pairs useless. Lockouts target repeated guesses against one account, which is not the pattern here. Spraying tries one common password across many accounts, and frequent forced changes are no longer recommended. Pass-the-hash is an internal Windows technique unrelated to this login pattern.
Question 3 of 10, Network Security
Large replication flows between two data centers pass through a Cisco Secure Firewall Threat Defense device. The flows are trusted and must bypass Snort inspection entirely, at the earliest possible point, to avoid added latency. Which configuration meets this requirement?
- AAn access control rule with the Trust action placed at the top of the policy
- BAn access control rule with the Allow action and no intrusion or file policy attached
- CA prefilter policy rule with the Fastpath action matching the replication flows Correct answer
- DAn access control rule with the Monitor action that logs the replication flows
Answer and explanation: Prefilter rules are evaluated before traffic is sent to Snort, and the Fastpath action keeps matching flows entirely in the data plane, so they never reach deep inspection. A Trust rule is part of the access control policy, so initial packets still go to Snort for rule evaluation before inspection stops. An Allow rule without intrusion or file policies still sends traffic through Snort for discovery. Monitor only logs and lets traffic continue to later rules.
Question 4 of 10, Network Security
A company wants its threat defense device to inspect inbound HTTPS traffic to its own public web server. The security team has the server's certificate and private key. Which decryption rule action is appropriate?
- ADo Not Decrypt with logging enabled at the end of each connection
- BDecrypt - Resign using an internal CA certificate
- CBlock with Reset for sessions that use TLS 1.3 only
- DDecrypt - Known Key using the server's certificate and key Correct answer
Answer and explanation: Decrypt - Known Key is used for traffic to servers the organization controls, because the device can use the server's own key material. Decrypt - Resign is meant for outbound traffic to external sites, where the device re-signs the server certificate with an internal CA that clients trust. Do Not Decrypt leaves the traffic opaque, and blocking TLS 1.3 sessions would break legitimate visitors.
Question 5 of 10, Network Security. Choose 2.
A security audit flags trunk-related VLAN hopping risks on campus switches. Which two measures mitigate VLAN hopping? (Choose two.)
- AConfigure user ports as static access ports and disable DTP negotiation Correct answer
- BEnable PortFast on access ports so they skip the listening state
- CSet the trunk native VLAN to an unused VLAN that carries no user traffic Correct answer
- DEnable DHCP snooping on all user VLANs and trust only the uplinks
- EEnable BPDU guard on every user-facing access port in the campus
Answer and explanation: Switch spoofing relies on DTP forming a trunk with an attacker's port, so static access mode with negotiation disabled stops it, and double tagging relies on the attacker's access VLAN matching the native VLAN, so an unused native VLAN (or tagging the native VLAN) stops that. BPDU guard protects STP, DHCP snooping stops rogue DHCP servers, and PortFast only speeds up port forwarding.
Question 6 of 10, Cloud Security
An organization needs to map application dependencies between thousands of workloads across on-premises data centers and public cloud, and then enforce allowlist microsegmentation using each host's native firewall. Which Cisco solution fits?
- ACisco Multicloud Defense with gateways inserted in each cloud VPC or VNet
- BCisco Secure Workload Correct answer
- CCisco Secure Firewall Threat Defense Virtual deployed at every data center edge
- DCisco Secure Access with Secure Private Access policies for each application
Answer and explanation: Secure Workload uses agents and telemetry to build application dependency maps and pushes allowlist policy to host firewalls such as iptables or Windows Firewall, which gives consistent microsegmentation everywhere. Multicloud Defense secures traffic through cloud network gateways rather than enforcing on each host. Edge firewalls do not control east-west traffic between individual workloads. Secure Access governs user access to applications, not workload-to-workload flows.
Question 7 of 10, Cloud Security
Workloads run in AWS, Azure, and Google Cloud. The security team needs egress FQDN filtering and intrusion prevention for VPC and VNet traffic, with one policy across clouds and gateways that scale automatically. Which Cisco solution meets these requirements?
- ACisco Secure Access
- BCisco Multicloud Defense Correct answer
- CCisco Secure Workload
- DNative cloud security groups
Answer and explanation: Multicloud Defense deploys auto-scaling gateways managed from a single SaaS controller and enforces ingress, egress, and east-west policy, including FQDN filtering and IPS, across major clouds. Secure Workload enforces host-based microsegmentation rather than gateway inspection. Secure Access protects user traffic, not workload egress. Native security groups filter by IP and port and do not provide FQDN filtering or IPS.
Question 8 of 10, Secure Service Edge
Cisco Secure Access must block uploads of documents that contain payment card numbers to unsanctioned cloud storage while allowing uploads to the corporate tenant. Which configuration is required?
- AA firewall rule in Secure Access that blocks outbound TCP 443 to the storage provider's ranges
- BAn API-based DLP rule that scans files already stored in the unsanctioned storage service
- CA DNS policy that blocks the cloud storage category for every roaming computer identity
- DA real-time DLP rule on web traffic with HTTPS decryption enabled for those destinations Correct answer
Answer and explanation: Real-time DLP inspects uploads inline through the secure web gateway, which requires decryption to see file content, and can block only sensitive data to specific destinations. Blocking the whole category at DNS also blocks legitimate use. API-based DLP works on sanctioned tenants the company controls, not on unsanctioned services. Blocking port 443 to a provider cannot tell sensitive from harmless content.
Question 9 of 10, Endpoint Protection and Detection
An analyst identifies a malicious file by its SHA-256 hash. The file must be quarantined wherever Cisco Secure Endpoint finds it on managed computers, even though the cloud does not yet classify it as malicious. What should the analyst configure?
- ACreate an advanced custom detection signature for the malware family
- BAdd the hash to an application blocking list assigned to the policies
- CAdd the hash to a simple custom detection list assigned to the policies Correct answer
- DAdd the file's callback address to an IP block list in each policy
Answer and explanation: A simple custom detection list contains SHA-256 hashes that Secure Endpoint treats as malicious, so matching files are detected and quarantined. An application blocking list prevents an executable from running but does not quarantine the file. Advanced custom detections use signature syntax for families and are unnecessary for a known hash. An IP block list controls network connections, not the file.
Question 10 of 10, Network Access, Visibility, and Enforcement
A SOC uses several Cisco and third-party security products. Analysts want detections from endpoint, network, email, and identity tools correlated into prioritized incidents with guided response actions. Which Cisco solution provides this?
- ACisco ISE
- BCisco Security Cloud Control
- CCisco Secure Workload
- DCisco XDR Correct answer
Answer and explanation: Cisco XDR ingests telemetry from multiple security tools, correlates related detections into incidents, prioritizes them by risk, and offers response actions and automation. ISE handles network access control. Secure Workload provides workload microsegmentation. Security Cloud Control manages security device policies rather than correlating incidents.